Digitising of photogrammetric instruments for cartographic applications

Photogrammetrta - Elsevier Publishing Company, Amsterdam - Printed in The Netherlands

DIGITISING OF PHOTOGRAMMETRIC CARTOGRAPHIC APPLICATIONS 1

INSTRUMENTS

FOR

G. PETRIE UniversiO, o/ Glasgow, Glasgow, Scotland (Great Britain)

(Accepted for publication October 26, 1972) ABSTRACT Petrie, G.. 1972. Digitising of photogrammetric instruments for cartographic applications. Photogrammetria, 28(5):145-171. In the first part of the paper, an analysis is made of the requirements of the various components of a photogrammetric digitising system designed for cartographic applications. The resolution, density and rate of measurements are established and the various possible measuring devices and systems inspected in terms of their suitability, both for the digitising of purely planimetric detail, e.g., on orthophotographs, and on different types of stereoplotting machine. The view is also expressed that new stereo-plotters might be devised, which are optimised for digitising rather than for graphical plotting as at present. The second part deals with the operational problems likely to be encountered, including stereo-compilation procedures, recording modes, feature codes, and the relative merits of pre- and post-plotting completion procedures. The possibilities of the on-line connection of stereo-photogrammetric digitising systems to a computer for control and preliminary editing purposes are discussed at some length. Finally some examples of photogrammetric digitising carried out both on experimental and on production bases are described and discussed. INTRODUCTION T h e question of digitising p h o t o g r a m m e t r i c m e a s u r e m e n t s b e c o m e s increasingly m o r e i m p o r t a n t as c o m p u t e r s are b r o u g h t into use at an ever greater n u m b e r of stages in the surveying a n d m a p - m a k i n g processes. Originally, in the late 1950's, p h o t o g r a m m e t r i c digitising was i n t r o d u c e d for aerial t r i a n g u l a t i o n w o r k to s p e e d up a n d to r e m o v e the errors inherent in the process of r e a d i n g plate or m o d e l c o - o r d i n a t e s visually. Shortly afterwards, it was used in a similar way for c a d a s t r a l w o r k a n d for the g e n e r a t i o n of digital t e r r a i n m o d e l s in engineering work. I n these latter applications, p o i n t s were still m e a s u r e d i n d i v i d u a l l y and in a n e a r - s t a t i o n a r y m o d e so t h a t the digitising a n d r e c o r d i n g processes r e m a i n e d basically the s a m e as for aerial triangulation. H o w e v e r , since the last I.S.P. C o n g r e s s in 1968, with the progress in the h a n d l i n g of e n o r m o u s quantities of d a t a in relatively inexpensive c o m p u t e r s , a n d the emergence, in several countries, of a c o n s i d e r a b l e a n d successful d e g r e e of a u t o m a t i o n of the c a r t o g r a p h i c p r o c e s ses which follow p h o t o g r a m m e t r i c m e a s u r e m e n t s , the question of digitising the p h o t o g r a m m e t r i c m e a s u r e m e n t s m a d e during the plotting of t o p o g r a p h i c line I Invited Paper for Commissions II and IV, International Society of Photogrammetry, Ottawa Congress, July 1972.

146

G. PETRIE

detail has become an important one, and one which will be intensively researched into and developed over at least the next decade. In a previous paper (Petrie, 1971), the author has discussed in detail the various advantages which may be expected to accrue from such a process. They include the preparation of topographic series at successively smaller scales, with a selection and some generalisation of what shall be plotted at each scale; the elimination of the re-measurement of the whole or a substantial amount of the plot inherent in the generation of areas for cadastral or taxation purposes; the creation of a geographical data base for data banks, planning organisations, public utilities (supplying electricity, water, sewage, telephone services), etc. When such digitising is being considered, the amount of digital data which will be generated is enormous and, again, estimates of this were made for various map scales in the previous paper. A major problem is the volume of data which would be generated at the large scales (1 : 1,250 to 1 : 10,000) which constitute the basic mapping scales of highly developed countries such as the United Kingdom, West Germany, Sweden, etc., and which call into question whether digitising should only be carried out at smaller scales. To do so, however, would mean that the smaller-scale maps could not be generated from the larger using data processing techniques, which is one of the main reasons for attempting to digitise in the first place. Also, the potentially large demand for digitised data from planning and public utility agencies, which will help to pay for the cost incurred in adopting digitising and automatic plotting, will be primarily for the urban and other intensively-developed areas, which are those covered by the large-scale basic series. However, it is probable that the problems regarding the amount of data being generated and the problems of storing and handling it will grow less with technological advances in storing data in compact form and with greater experience in processing it. Arising from the above problems is the view that the photogrammetrist should not attempt to digitise data for cartographic purposes and that he should continue to produce his graphical plots in the normal way or perhaps with special colour coding for the cartographer to digitise at a later stage for his own operations. Since this would involve an expensive, time-consuming re-measurement of the whole plot with a possible deterioration in accuracy, and cartographers already have the great problem on. their hands of digitising the huge amount of existing cartographic material, it is difficult to see the logic of this view, except to agree that it would certainly be a way of not having to face the difficult technical and operational problems of photogrammetric digitising. However, if the photogrammetrist's claim to a stake in the digitising field can be upheld it is more difficult to define the point where this interest should end. The digitising of point and line detail in the stereo-plotter is only one form of input to the cartographic process: much additional information comes from existing maps which have been converted to digital form by the cartographer and from details collected in the field by the topographer. All this different information

PHOTOGRAMMETRIC DIGITISING FOR CARTOGRAPHIC A l MS

147

has to be merged together, compiled, edited and generalised before being plotted out in the form of colour-separated manuscripts--tasks which are the traditional preserve of the cartographer. A question which will inevitably arise is how far should the photogrammetrist become involved with these cartographic problems? No doubt the demarcation line will vary from organisation to organisation, but one can see already that in organisations which are concerned principally with engineering-type applications, e.g. Highway or Public Works departments, usually the amount of compilation, editing and generalisation is small, the final product may not be a multi-coloured map and a principal aim is the simple replacement of hand drawing and scribing by automatic plotting, so that the photogrammetrist can become involved to a limited extent with these traditional cartographic problems. However, in national mapping agencies, where a whole series of topographic maps at different scales and with differing degrees of generalisation may be produced from the photogrammetric plots, then the cartographic problems loom much larger and it is more difficult to justify the intrusion of the photogrammetrist into this area. However, as well as considering the limits of the photogrammetrist's interests, one must recognise that, just as with the production of photomaps, the introduction of digitising of photogrammetric plotting will bring the photogrammetrist and cartographer into ever closer contact and a higher degree of cooperation and integration with a better understanding of each other's problems and procedures is a necessity for successful implementation of digitising, data manipulation and automatic plotting for topographic mapping. Having discussed some of the general problems arising from photogrammetric digitising operations, the remainder of this paper will be divided into two parts: (l) Technical problems associated with the provision of digitising equipment for topographic photogrammetric operations. (2) Operational problems, possibilities and experience. This division approximates to the respective areas of interest of the two I.S.P. Commissions (II and IV) to whom this paper has been given. TECHNICAL PROBLEMS ASSOCIATED W I T H DIGITISING EQUIPMENT

When considering recent technical developments in digitising photogrammetric measurements, one must first note some general points. First of all, the hardware requirements for digitising photogrammetric measurements for later cartographic plotting are often quite different to those which have been satisfied by the first- and second-generation digitising equipment developed and used in the late 1950's and throughout most of the 60's, as typified by the Wild EK-2, 3 and 5, the Zeiss Oberkochen Ecomat 1, the Zeiss Jena Co-ordimeter and similar systems. With all of these, one could satisfy the digitising requirements for aerial triangulation, where perhaps ten individual points per stereo-model need measurement, and for the several hundred points per model required in cadastral work

148

G. PETRIE

or engineering digital terrain models. In topographic plotting work, one has to plan for the digitising of tens of metres of line plotting per model, which, depending on the density of recording, means the measurement and recording, at speed, of from several tens of thousand to several hundred thousand points per stereomodel. Obviously, there has been and still is a need for new digitising equipment to meet these quite different requirements. Until recently, most photogrammetric instruments have had digitisers manufactured and attached by the instrumental manufacturer. A feature of the last few years has been the widespread installation or attachment of systems made by specialist digitising firms, often with little or no connection with the photogrammetric manufacturers. Perhaps this has been due partly to the shortcomings of the equipment offered by these manufacturers, but it also results from the fact that the devices and systems which have been developed in many related fields (e.g., numerical control of machine tools, non-photogrammetric cartographic work, etc.) are readily adapted for installation on photogrammetric equipment. Closely associated with the now-widespread use of non-factory supplied digitisers is the use of local firms. For example, SAAB (Sweden), Instronics (Canada), Auto-trol and Dell Foster (U.S.A.), d-Mac and Faul-Coradi (U.K.) have all supplied digitisers on Wild A8 machines installed in these various countries. The reasons given for this trend are sometimes technical (superior performance), sometimes commercial (cost, especially when shipping and customs charges are added to ex-factory prices), sometimes operational (ease of servicing), sometimes political (the wish to use local manufacturers), or a combination of these factors. In the remainder of this part, an analysis of the various components and of photogrammetric digitising systems will be made under the following headings: (A) Resolution, speed and density of photogrammetric measurements for cartographic purposes; (B) general principles and methods of measurement which meet these requirements; (C) x, y co-ordinate measuring systems; (D) x, y, z coordinate measuring systems; and (E) recording and storage devices. A fairly analytical and generalist point of view has been taken in each of these sections since the intention of this paper is to focus attention on the digitising problem as a whole, rather than simply attempt a survey of devices presently offered on the market.

Resolution, speed and density of measurements lor cartographic purposes As already mentioned, it is important for designers, constructors and operators of digitising systems to realise that the specifications and performance required from these for cartographic purposes are often quite different to those demanded for the aerial triangulation, cadastral and engineering work which have been the chief tasks for which digitising equipment has been used till now. Resolution. In aerial triangulation work, measurements on photographs with eomparators or in the model with stereo-plotting equipment need a resolution in

P H O T O G R A M M E T R I C DIGITISING F O R C A R T O G R A P H I C AIMS

149

the range 1-5 um; in cadastral and D T M work, resolution requirements are in the range 5-20 ,urn. However, the many measurements needed for cartographic work will be made during continuous or semi-continuous plotting of mainly linear features; and so resolution requirements will be less than when only a few very precise point features (signalised points, drilled holes, reseau crosses, etc.) have to be measured and more time can be devoted to the measurement of each individual point (as in aerial triangulation operations). Remembering, too, that the final output will be plotted graphically in an automatic co-ordinatograph, then one may assume that a resolution of 25 !~m in the stereo-model will be more than sufficient for the purpose.

Density. This leads directly to consideration of the density of measurements which is needed for cartographic digitising. In aerial triangulation and other similar work, only a comparatively few points need to be measured with a relatively long time interval between them. Slow recording and output devices without mass storage facilities (e.g., paper tape or card punches) are quite appropriate. For cartographic work, the density is much higher and the interval between measurements very short, with consequent effects on the measuring/recording systems. From previous papers (Howard, 1968; Uhrig, 1970; Petrie, 1971) it seems that the density of line detail likely to be found in topographic maps lies in the order of 2-12 cm/sq.cm, depending on the scale of plotting and on the type of terrain being considered: an additional, comparatively trivial number of point features will also have to be measured. Considering the line features, these can be divided into two main groups. The first comprises straight-line and mainly man-made features--buildings, certain physical boundaries (e.g., fences and walls), etc.--where only the beginning and the end points of such features (plus perhaps one or two intermediate points) need to be recorded. This is done in conjunction with a suitable coding or instruction which can permit later plotting of the continuous straight lines via the hardware or software of the automatic co-ordinatograph. This reduces both the number of measurements to be made and the amount of storage to be provided in the recording device or in the computer. The second group comprises the more sinuous linear features--contours, rivers, forest and field boundaries, many roads and tracks, etc.--to be plotted. These features are at a maximum in areas where man's activity is least and at smaller scales. Obviously, with these lines, many more points have to be measured and recorded if an accurate plot is to result later; the actual density depends on the characteristics of the measuring/recording device itself; the scale of the finally plotted maps; and the suitability of the algorithm which is adopted for the later reconstitution of the lines during plotting. Speed.

This factor is closely related to the density of recording. Again, it is a matter which is very different when digitising for cartographic operations than

150

G. PETRIE

for aerial triangulation. With the latter, a very slow final movement is made when measuring the actual triangulation point, but unfortunately this is coupled with a requirement for very rapid rates of movement between these few points. This, when associated with the high resolutions which are needed for aerial triangulation work, imposes great strain on the counting (and decoding) capabilities of a digitising system. In this respect, the designer must also allow for the considerable accelerations which may occur due to accidental bumping or pushing of the measuring carriage, particularly in a free-moving type. If a deliberate or accidental rate of movement of 0.5 m/sec is achieved, a counting speed of 50 kHz will be generated in a measuring system which resolves 10 urn, and 500 kHz with a 1 !~m resolution. When digitising for cartographic operations, the designer's problems are lessened in some respects because of the lower resolution requirements. Also, the normal speed of plotting linear features is not too high, 1.5--2.5 mm/sec, though allowance has still to be made for accidental displacements. However, even at these plotting speeds, high-density digitising at the closest practicable interval of 0.1 mm (100 um) will require 15 to 25 points to be measured and recorded per second. This is, of course, a data acquisition rate which could not be accommodated by first- and second-generation digitising equipment where the coordinates of a single point took from 2 to 12 sec to be measured and recorded, and measuring operations could not be performed "on the fly". Summarising, one may say that, while requirements for resolution and rates of movement are lessened when digitising photogrammetric measurement for cartographic purposes, the requirements for the rate and density of information to be acquired increase many fold, so that high-speed measuring, decoding and recording have to be adopted. This has been made possible by the introduction of opto-electronic measuring devices, integrated circuit logic and high-speed, highdensity, recording and storage devices such as magnetic and high-speed paper tape recorders. However, if a digitising system has to be supplied for both aerial triangulation and cartographic work, then it is obvious that the quite differing needs for high resolution, high rates of movement, high density of recording, etc. can stretch the digitising system to the present limits of the art.

Measurement possibilities In considering digitising systems which will meet the requirements defined above, attention will first be given to the basic characteristics of the measuring devices themselves before inspecting their capability and potential when mounted on a photogrammetric instrument. Absolute and incremental methods of distinguish between these two forms of suitably coded linear scale, will record for each position in the photograph

measurement. First of all one must measurement. An absolute device, e.g., a a unique value for each co-ordinate axis or stereo-model area occupied by the

P H O T O G R A M M E T R I C DIGITISING F O R C A R T O G R A P H I C AIMS

151

measuring device (space rod intersection point, measuring mark or cursor, etc.). Every other position will be distinguished by having a different and unique value. If the measuring process is interrupted, e.g., if the power which ac::uates the device were to be cut off, then, on resumption, the same value as before will still be measured and recorded on each axis for every point in the photo or model. With an incremental device, positions are not co-ordinated in terms of fixed values for each co-ordinate axis, but in terms of the total count of distance increments made from an initial starting point, which may be specified to be anywhere within or outside the photograph, map or model area. Obviously even the smallest interruption in detection or counting of increments in such a device will lead to quite erroneous co-ordinate values for every position measured thereafter. While the absolute system has the obvious feature of constancy of value, irrespective of power interruption or failure, it does require a large number of tracks or channels to provide a unique code or value for every position, particularly if both a high resolution and a long measuring distance are to be provided. For example, if a 10-!~m resolution is to be provided over a 70-cm distance (as along the y-axis of a stereo-plotting machine) then some 70,000 (i.e., 2 l';) values have to be specified uniquely. Using a binary-based coding system, suitable for electromechanical or electronic detection of values, then 17 tracks or channels must be provided on the scale and at least one detector or counter per track. Furthermore, such a scale is more difficult and expensive to manufacture and complex decoding and combining of the binary signals must be provided. Also it is difficult, though not impossible, to key in a predetermined value for a particular point, as is desirable in some cartographic operations. By contrast, an incremental system has a much simpler scale, e.g., a simple grating may be provided or one with an additional index grating as in the wellknown moir6 fringe system. Either system is less expensive than the corresponding fully-coded absolute scale. Essentially, the measured value is produced by counting the series of increments or pulses from a starting value. To achieve the correct value, the direction of movement has to be sensed and vibration in particular has to be guarded against or values will be added to the count for no change in position. The effects of deliberately or accidentally cutting off the power supply or of miscounting in an incremental type of device have already been mentioned. For all that, the tendency with most recently designed or applied photogrammetric digitisers is to adopt incremental techniques, though the absolutely coded linear scale used by Sopelem for the Pressa 226 is a notable exception. The main advantage of the incremental systems is the lower cost, not only in the manufacture of the scale, but in having to provide only two or four channels or tracks for detection and counting of the increment. In this respect, the extreme reliability in counting which is possible with current solid-state electronics technology has played a considerable part in the adoption of incremental devices in photogrammetric practice. A further point is the ease with which a given or convenient value can be entered in an incremental system.

152

c. PETRIE

Physical principles and methods of measurement. Turning to the actual digitising devices which could be attached to photogrammetric machines, these are legion in number and vary enormously in principle, resolution, range, speed, etc. Not all those which could be attached meet photogrammetric requirements, e.g. there are numerous devices used in the mechanical engineering field which give an appropriate resolution, but only over a relatively short distance, so that the resolution (which is often a fixed proportion of the length measured) is insufficient when used over the distances encountered in stereo-plotters. Looking at yet another possibility, intefferometric methods of measurements were limited, till recently, to short distances and to slow counting speeds due to the inadequacies of the light sources used. With the advent of the laser as a suitable highly coherent and intense light source, interferometric methods can now be considered as a possibility for photogrammetric digitising. However, the resolution of the method is almost too high--if individual fringes are counted, resolution is automatically 2/2 = 0.3 , m and techniques exist which can result in 2/8 = 0.08 , m being resolved. Furthermore, the cost of a two- or three-axis laser interferometer is still very high, so that such a device is more appropriate to calibration work (i.e., to the testing of comparators, stereo-plotters and co-ordinatographs, and their associated measuring systems) than to its everyday use in the photogrammetric plotting room. Inspecting still other possibilities, there are a great number of electrical methods which measure resistance, capacitance or inductance as a function of the distance traversed by a moving carriage, but, as yet, these have not been applied to photogrammetric work. There is a need for these numerous devices to be inspected systematically to ascertain whether they are appropriate for photogrammetric purposes now that there is a potential need to digitise almost al! photogrammetric measurements for later data processing. Turning to the methods which have actually been used or show obvious potential for use, they can be divided into three groups: (a) those which employ linear scales for measurement; (b) rotary encoders; (c) area methods which employ a grid of sensors which are usually an integral part of the support or measuring surface of an instrument. Linear scales have been used on stereo-plotting machines for a considerable time for visual measurement of x and y co-ordinates, e.g., with metal scales on the S.O.M. Poivilliers Stereotopograph B and utilising glass scales on the GalileoSantoni Stereocartograph IV. They can equally readily be used for electronic digitising systems and are available in both absolute and incremental forms. The absolute type of multi-track coded scale is available from firms such as Heidenhain (West Germany) and Rank Precision (formerly Hilger and Watts, U.K.). The simpler, incremental type using gratings has also been developed extensively for industrial use, particularly those which employ some type of index or reference grating such as the moir6 fringe system of Ferranti (U.K.), the Heidenhain (West Germany) LIDA system (Fig.l) and those developed by Dynamics Research Cot-

153

PHOTOGRAMMETRIC DIGITISING FOR CARTOGRAPHIC AIMS

poration (U.S.A.) and by Philips (Netherlands). With these, the sensed portion of the scale is not a single line but the average of hundreds of adjacent lines, thus reducing the effect of graduation errors.

Electronics

Glassindex

Lamp

Photo-cells

.-

Coded measuringscale

Coupling

Photo cells

Rotary glass disc gratmg

with

Lillear Absolute Measuring System

Rotary Incremental Measuring System

Glgratin! assirmde~ Steelwithmeasuri scal ng e ~rating Zero

Linear Incremental Measuring System

g r a t l ~ Counter DIAGRAMS OF THREE MEASURINGSYSTEMS OF HEIDENHAIN ( W.GERMANY )

Fig.1. Diagrams of three measuring systems of Heidenhain (West Germany).

Still other linear systems have been used in photogrammetric instruments. One, used in the Keuffel and Esser Ecars system mounted on Kelsh Plotters, employs precision wire-wound potentiometers, with the wires wound around a bar. A d.c. voltage is applied along the wires, while a sliding contact connected to a digital voltmeter is attached to the carriage and bears on them. The voltage read is linearly proportional to the distance travelled by the carriage. Another system (proposed at the I.S.P. Congress, Lausanne, 1968, for use with the Galileo-Santoni Stereocomparator) is the Farrand Inductosyn, which has stator and rotor wires imprinted as metallic conductors on glass scales, and gives a distinct set of stator voltages for any given rotor position. Looking at the materials used for linear scales, glass has been universally popular, particularly if high accuracy has been required. Also it lends itself to the use of high-speed, opto-electronic methods of pulse detection and counting. For comparator work, where the travel is comparatively short, glass scales are excellent, but they are not so easy to manufacture in the lengths encountered in the stereo-models of plotting machines, and are correspondingly expensive. An

154

c. PETRIE

important recent development therefore has been the use of a grating of highly light-absorbent, black lines etched into the reflective surface of a stainless-steel tape which is supported by a steel carrier frame. Such a grating manufactured by Ferranti has been used in the Faul-Coradi digitising system recently attached to the Wild A8 Autograph of the University of Manchester (U.K.). Rotary encoders have been widely used in photogrammetric digitising systems. As far as the measurements are concerned, the same physical principles and devices have been utilised in rotary shaft encoders as in the purely linear measuring systems already discussed: absolutely coded glass or metal discs, incremental gratings, wire-wound potentiometers, the Inductosyn system, etc. The pulse detection, counting, decoding and storage requirements and systems are substantially the same whether a linear or rotary system is used and frequently these are interchangeable in modular designs. However, with all rotary devices, since ultimately a linear distance has to be determined, it is necessary to have some linear to rotary converter mechanism, e.g., a lead screw and nut, or a precision rack and pinion, so that the distance can be measured in terms of the number of turns and partial turns made by the rotary encoder. The angle which can be resolved by such an encoder is limited, the actual amount varying with different units depending on the design principle. Resolution can be increased by ensuring that the rotary encoder is turned several times for each revolution of the driven component, but gearing inaccuracies place a limit on the resolution which is achievable. A disadvantage of all rotary measuring systems employing mechanical conversion or scaling elements is that the elements themselves have inherent errors and are also subject to wear. By contrast--and especially if a non-contacting system is u s e d - - a linear scale has complete freedom from such errors and for precision measuring systems they are increasingly used. The disadvantage of the linear type is the slightly higher cost of these scales as compared with the rotary discs. This cost differential is least with the short distances utilised in photogrammetric comparators and greatest with the longer distances and the three axes which need to be measured in the stereo-plotters used for cartographic digitising. However, the extra cost of using linear scales is only a small proportion of a complete digitising system and must be balanced against the advantages already mentioned. The third and final group has been variously described as the area-, gridor mat-type of digitiser. The origin of these devices may be discerned in the Rand Tablet, developed by T. O. Ellis of the Rand Corporation, in which a series of parallel electrical conductors in both the x and y co-ordinate directions is embedded in a base or matrix. Attached to the tracing/measuring device (which can be moved freehand over the mat or tablet) is a coil which generates an electromagnetic field which can be picked up by the array of conductors below. This generates a series of pulses which indicate the x and y co-ordinate values; if the measuring point is moved to a new position, a different set of pulses will be

PHOTOGRAMMETRIC DIGITISING FOR CARTOGRAPHIC AlMS

~55

generated to produce new co-ordinate values. The original devices which were developed on this principle, e.g., the Sylvania Tablet and the Data Equipment Corporation Graphacon, had limitations both in the area covered (25 7< 25 cm) and in resolution (0.25 mm -- 250 , m ) , so that they had little application to photogrammetric or cartographic operations. However, with the recent development of higher-resolution mats of great area (up to 1.5 X 1.5 m), particularly by the Bendix Corporation (Bailey et al., 1969: Hattaway, 1971) the situation has been transformed. The physical speed of movement of such a device is practically unlimited and, since there are no mechanical moving parts (except the cursor and its coil), so wear is at a minimum. The resolution is now 25 ,um and the accuracy m x -- m y = + 50/~m, which means that, while the system is not applicable to comparators and other high accuracy photogrammetric machines, it is well suited to digitising for cartographic purposes. Furthermore, it has the merit of being easily mounted on those stereo-plotting machines, e.g., the Zeiss Oberkochen Planimat, DP-1, Kelsh Plotter, Wild B8, Kern PG-2, which feature a fiat-bed measuring/plotting surface. The electronics needed are the same as those developed for the other types of pulse-counting, incremental measuring systems. x, y Co-ordinate measuring systems

~fhese systems can be divided into two groups. (1) The first comprises those incorporated in instruments such as monocomparators, which can achieve measurements of high resolution and accuracy and are usually concerned with analytical, numerical work for calibration and control purposes. Since this paper is concerned with photogrammetric digitising for cartographic purposes, these have been excluded from consideration. (2) The second group includes devices or systems, e.g., cartographic digitisers, which permit measurements of a lower resolution and accuracy to be made on single photographs and recorded for later processing for topographic cartographic purposes. Typical of these operations are the measurement of planimetric data from orthophotographs; the measurement of revision material (via rectification) from single aerial photographs; and the use of parallax-type instruments for heighting information. Some may question whether these operations are strictly photogrammetric; since they involve positional measurements on photographs for topographic purposes they have been included in this paper. The cartographic digitisers of group 2 have been intensively developed over the last decade, along three main lines: the lock-on, line-following type of device in which a photo-electric sensing head follows the lines on an existing line map or colour-separated document; the area scanner in which similar material is systematically scanned, usually in a raster pattern, over its whole area for the presence or absence of lines; and the manually- or operator-controlled type of digitiser where an operator measures the individual positions, lines or photograph detail required for map construction.

156

G. PETRIE

The first two types seem, at present, to have only a limited application to topographic photogrammetric work. The lock-on, line-following type of device has been used by the U.S.G.S. for the digitising of graphically-recorded profiles measured on a Kelsh Plotter for the later off-line production of orthophotographs (Hughes et al., 1971). The area scanner can utilise a C.R.T. to produce the raster scan, but more usually it involves a drum around which the existing plot, compilation, map or photograph is wrapped for scanning. For aerial photographs, the main use of drum scanners has been to acquire densitometric information for photo-interpretation; photogrammetric information which can be used directly for cartographic purposes is not yet extractable. This brings one to the third class: the manually- or operator-controlled type of equipment, which is of much greater interest to the photogrammetrist. This group may conveniently be divided into two categories: (a) Those in which the tracing/measuring device is mounted on a crcssslide or double carriage arrangement which readily allows the measurement of x and y co-ordinates. An enormous number of manufacturers currently offer this type of device, with great variations in specification and performance in respect of resolution and accuracy, and consequently in cost also. At the lower end of the performance and price range are those devices which utilise modified engineer's draughting machines for the cross-slide system: at the upper end of the range, modified precision co-ordinatographs (e.g., the Coradi Coradograph) or speciallybuilt cross-slides as built for the d-Mac Cartographic Digitiser and Dell Foster Graphic Quantiser provide the necessary motions. Although linear gratings are ideal for the generation of the co-ordinate values, generally, even the highquality types utilise a linear/rotary conversion, either via a wire and pulley arrangement or via a precision rack and pinion set-up, and rotary encoders. A development of the cross-slide type is that where the slides are mounted below the support table for the photograph or plot. Measurement is then made with a free-moving stylus or cursor around which is placed a field coil, whose signals are picked up by the sensor coils of the detector head mounted on the cross-slides (Fig.2). Through the power of two servo motors, this head is automatically kept in position directly below the stylus. The x and y co-ordinate

Map

or

Orthophotograph

ing stylus with A.C. field coil p~ck up coils

support surface

driving wire support carriage

"

Fig.2.

y

)-

For explanation, see text.

PHOTOGRAMMETRIC DIGITISING FOR CARTOGRAPHIC AIMS

] 57

positions of the detector head are then digitised via rotary encoders as in the nonservoed versions. This type of servo-driven digitiser, which has been developed principally by d-Mac as the Pencil Follower, and also by Instronics, as the Gradicon, has the advantage of leaving the photograph or map surface completely clear as compared with the unservoed type. The higher-accuracy types of cross-slide digitiser generally provide a resolution of 25-50 u m (0.001-0.002 inch) and an accuracy of m x ~ m y ---- ±- 50 to 100 /~m, which is certainly adequate for photogrammetric-cartographic purposes. (b) The mat- or grid-type of digitiser is the type where the measuring device rests directly on the photograph, plotting surface or map and interacts with the positioning system, which is an integral part of the flat-bed support or base for the image. Usually a matrix of x and y position wires is embedded in a base of fibreglass or similar material as described above. Quite apart from the obvious general application of these mats to the measurement of detail and points from orthophotographs mounted on a light-table (translucent mats have been developed), the Bendix Datagrid in particular has been adapted for use in two other specific pieces of equipment. One is the LR-2 Line Rectifier (Forrest, 1972), built by Bendix for the U.S. Naval Air Systems Command, in which image positions are measured using the Datagrid cursor, the x and y co-ordinates being fed directly to a small computer which transforms and rectifies them to the corresponding terrain values, which can then be plotted by a small co-ordinatograph which is also controlled by the computer. A second device, which is currently under development by the U.S. Army Engineer Topographic Laboratories, involves the use of a digitised parallax bar used in conjunction with a mirror stereoscope, the corresponding x and y positions of each measured height point being given by a Datagrid mat. The measurements are again fed to a desktop computer which already has the rotation matrix, control point values, etc., stored for each pair measured from aerial triangulation, so that the terrain position and height of any additional point can be generated quite rapidly. x, y, z Co-ordinate measuring systems

It is but a small jump from the x, y measuring systems just considered to those such as our familiar photogrammetric stereo-plotting machines which measure along an additional third z-axis. In fact, many of the manually-controlled x, y devices already mentioned are offered and have been used directly on stereoplotting machines for digitising for cartographic operations. Considering the possible methods which can be used for digitising measurements made in stereo-plotting machines, they are conditioned by the mechanical arrangements made for controlling the x, y positions of the measuring point in the model space. From this point of view, they can again be divided into: (1) The cross-slide or double carriage type which can be further subdivided into: (a) those which employ handwheel control of the plotting/measuring motion via shafts or lead screws and nuts, e.g., the Wild Autographs A8 and A10;

158

G. PETRIE

and (b) those which employ a direct, freehand control of the plotting/measuring motion, e.g., the Galileo-Santoni Stereosimplex IIc and III series. (2) Those which employ a fiat-bed surface on which a mobile measuring device can be moved freely, e.g., the Kelsh Plotter, Balplex, Wild Aviographs B8 and B9, Kern PG-2, etc. The Zeiss Oberkochen Planimat and Planicart designs are exceptional in that they feature both a cross-slide arrangement and a flat-bed surface. In all the above categories, the third, z-motion, whether foot-, hand- or thumb-operated, involves a rotary movement via a shaft, so that this is not a variable from the digitising point of view. (la) With these stereo-plotting machines, the use of handwheels, shafts or lead screws makes them, at first sight, obvious candidates for the addition of rotary shaft encoders. However, while a few machines, e.g., the Zeiss Oberkochen Stereoplanigraph and Zeiss Jena Stereometrograph, only have such a handwheel/ lead screw/carriage system, most stereo-plotting machines in this category have the possibility of disengaging the nut from the lead screw to provide a rapid motion between points during relative orientation. Unfortunately, the use of rotary shaft encoders means that, during absolute orientation and the measurement or plotting of the stereo-model, the nut connecting the lead screw to the carriage must always be engaged or the count will be lost (with the notable exception of the Kern PG-3). This is a major disadvantage during aerial triangulation where a rapid movement between a few measured points is desirable. During digitising for cartographic purposes, this disadvantage is rather less pronounced. Obviously, the use of linear gratings would be of considerable advantage in that the count would always be implemented whether handwheel or freehand movements were being performed. The disadvantage in using linear gratings is that they are a little more expensive with length if large dimensions are involved as in stereo-plotting machines, whereas the cost of attaching rotary digitisers to this type of stereo-plotter is independent of the size of the model area to be measured. However, another important consideration regarding the use of mechanical conversion elements such as a lead screw and nut or rack and pinion is that these elements have inherent manufacturing errors and are also subject to wear with use. As previously discussed, the adoption of linear scales would overcome these difficulties. In the related fields of machine tool construction, scientific instrumentation, nuclear bubble-chamber photographic measurements, etc., where there is a similar need for high precision x, y or x, y, z measurements, the use of linear gratings is widespread. While there seems to be a growing realisation of the advantages of these gratings for photogrammetric measurements leading to their use on some recent comparator designs (e.g., the Kern MK, Dell Foster, Sopelem, etc.), they have as yet been used on only a few stereo-plotting machines in this group, e.g., on the Wild A8 of the University of Manchester, already mentioned, on the Sopelem Pressa machines, and on the Zeiss Oberkochen Stereoplanigraphs

PHOTOGRAMMETRIC DIGITISING FOR CARTOGRAPHIC AIMS

159

of the U.S. Army TOPOCOM digitised by Dell Foster. (lb) On the non-handwheel controlled machines typified by the GalileoSantoni designs, the advantages of the free and rapid movement during all modes of operation are obvious. In line with this, linear, glass and metal scales have been used for visual~ co-ordinate measurement since the introduction of these designs. This makes it all the more curious to see the mounting on such machines of special linear to rotary converters in the shape of precision racks and pinions with all the extra cost involved and the possibility of wear and resultant inaccuracy over a period of time. (2) In the fiat-bed type of machine such as the Kelsh Plotter, Balplex, Zeiss Oberkochen DP-1, Wild B8 and B9 Aviographs and Kern PG-2, the main point is the very wide variation in the area of the stereo-model to be digitised from the PG-2 (15 X 23 cm) to the Kelsh ( 4 5 ) < 100 cm). So far, with all of these machines, the main approach to digitising has been first to mount a double-carriage coordinatograph so that the plotting movements are readily split up into their x and y components, and then to implement a conversion of the linear motions to a rotary movement via racks and pinions, in order to convert them to the mechanical arrangement of group la above. This has been done both by the photogrammetric factories (e.g., on the Wild B8 via a Coradi or Wild co-ordinatograph and Wild EK-8 electronics) and by the independent producers of digitising equipment (e.g., the Kern PG-2 and Kelsh Plotter with Dell Foster digitising equipment). Again, one must question and doubt whether this is an optimal solution to the problem. The purchase and mounting of a high-precision co-ordinatograph equipped with racks and pinions adds considerably to the expense of the equipment while the additional weight and inertia are also disadvantages. For these machines, the use of a mat-type digitiser (which carl be supported by the flat-bed surface) in combination with the z movement being digitised by either a rotary or linear digitiser appears more logical and convenient than the present arrangements. The Bendix Datagrid has in fact been adopted for use on Zeiss Oberkochen Planimat and DP-1 Plotters and Kelsh-type stereo-plotters in the United States and one would expect increasing adoption of such digitisers for this class of machine in future. With the very large Datagrid mats which have been produced, it would be quite feasible to digitise the several adjacent models which can be set up on the Kelsh and other double optical projection plotters. One special point about the use of either the cross-slide or mat-type digitiser on flat-bed surfaces is that they are readily demountable to be used as a cartographic digitiser on existing maps or orthophotographs. Another possibility of acquiring x, y, z co-ordinates, which has not been considered as yet, is through analytical plotters such as the OMI-Bendix AP and AS series. The values actually measured in these machines are the x and y coordinate positions on the individual plates making up the stereo pair. These values, which are generated by rotary or linear digitisers, are fed as input to the realtime computer to generate the model co-ordinates which can position the plotting

160

c;. PETRIE

pencil on the machine's co-ordinatograph. The x, y, z values which are recorded on magnetic tape will normally be tapped from the computer output interface and there is no requirement to measure and digitise model co-ordinates as there is in an analogue type of stereo-plotter. Reco,'ding and storage devices

Measurements made for aerial triangulation, cadastral and engineering purposes are comparatively few in number and have a relatively large time-interval between each. These low densities and the slow acquisition rates of the data are reflected in the output devices--card and tape punches, teleprinters and typewriters--which have been and are commonly used on such operations. For cartographic digitisip_g, data acquisition rates are much higher. At normal line plotting speeds of 1.5-2.5 mm/sec, 15-25 points/sec need to be measured and recorded if positiors are measured at intervals of 0.1 ram: for profile scanning of orthophotographs with correlator-equipped stereo-plotters, rates of movement of up to 6 mm/sec can be expected, so even higher data-recording rates have to be planned for. Not only is the data generated at a high rate, but it is generated in enormous quantities. Depending on the type of terrain, the density of detail present and on the sampling rates which are being applied (a problem studied in detail in Petrie, 1971), one may expect between 20,000 and 200,000 individual points to be measured and recorded in a stereo-model. When one reckons that 10 bytes or characters are required to specify a single x, y position, then the storage capacity of punched cards or a paper tape reel (which is from 80,000 to 90,000 characters per 250 m reel, or 8-9,000 positions) is quite inappropriate to the p r o b l e m - - n o t least because of the problems of handling which result. From the above considerations, the use of a high-speed recording device utilising a high-density storage medium is obligatory for cartographic digitising. The most obvious medium is magnetic tape. On a single tape, some 10,000,000 characters or 1,000,000 individual positions may in practice be recorded on a single 720 m long, 7 or 9 track tape at 556 b.p.i. (bits per inch), at a practical recording speed of 30 x, y co-ordinate sets per second. Of course, the use of a write-only incremental magnetic tape recorder does increase the cost of the output recording device by a factor of two to three over a high speed paper tape punch (the keyboard or typewriter for generating headers being an item common to both devices), but this represents only a 25% increase in the cost of the whole measuring/ recording outfit. The advantages of noise-free operation and possible higher reliability are other gains from the adoption of magnetic tape. At the present time, only a very small proportion (perhaps 10% or less) of the stereo-plotters in a photogrammetric office are equipped with digitising equipment for aerial triangulation, eadastral, digital terrain model work, etc. If there is to be a big increase in this proportion with the adoption of digitising for cartographic work, then a very considerable financial investment will be needed

PHOTOGRAMMETRIC

DIGITIS1NG FOR CARTOGRAPHIC

AIMS

161

for purchase of this equipment. If a number of stereo-plotters have to be equipped for the purpose (as distinct from equipping a single machine for experimental work) then serious consideration has to be given on economic grounds to eliminating the need for multiple recorders and to use a suitable mini- or process-control computer as an output device. A suitable real-time computer can act as a scanner and sorter of the digitised data output from several stereo-plotting machines, allocating space and writing the data on a mass-storage device such as a magnetic" disc. This has already been done, e.g., with the Concord cartographic digitising tables of the U.S. A r m y TOPOCOM, which are interfaced to CDC 1700 computers; on the Wild B8 Aviograph connected to a PDP-11 computer at the Canadian Federal Survey and Mapping (Harris, 1971); and in certain French agencies as attachments to SFOM-Matra stereo-plotters. If such a solution is adopted, the way is open to a still more sophisticated operation, since the computer may also be used to monitor and control the digitising operation and edit the digitised information. Of course, all digitised data, e.g., from a magnetic tape recorder, needs such verification and editing before it is in a condition to be used for compilation or plotting. Normally this is done offline on a computer in batch mode. However, the direct interfacing of a computer on-line to several stereo-plotters offers the prospect of the editing being done and mistakes corrected while the photographs are still in the stereo-plotters, a matter which will be discussed in more detail in the next section of this paper. Whether it is economic to do so is an open question. Protagonists of on-line and off-line verification and editing procedures exist already and one hopes that quite some light, as well as heat, will emerge from their discussion at the Ottawa Congress.

Summary One can reasonably assume that large-scale retrofitting of digitisers to existing stereo-plotters will take place if the present efforts into automatic cartography are successful. F r o m the discussion above, one can see that there is a real need for a more systematic approach to the design of suitable digitising/ recording systems for such plotters. Furthermore, one may also assume that new stereo-plotters need to be designed and developed which are optimised for digitising operations from the outset. These might be very different to current designs: e.g., the present high-precision graphical plot could be replaced (with a considerable financial saving) by a less accurate plot which simply allowed the operator to see if a point or line feature had been plotted or not. Alternatively, a quick-look interactive display, such as a storage cathode-ray tube might be considered, especially if a computer is being used as the output-cure-editing device. Considerable attention also needs to be given to the ergonomic side: better keyboards and co-ordinate displays need to be provided so that the loss of time in entering headers and alphanumeric information is minimised. Other ideas which might be investigated include the simple display of co-ordinate values (transformed

162

C. PEVRIE

if necessary in the control computer) in the operator's viewing oculars for the rapid setting of the measuring mark on pre-set or pre-determined positions for digital terrain models, signalised points, etc. No doubt, some of these ideas will be found in practice to be inappropriate or too expensive to implement: but there can be no doubt either that some radical re-thinking of stereo-plotter design will accompany any widespread acceptance of digitising for cartographic purposes. OPERATIONAL

PROBLEMS"

POSSIBILITIES

AND E X P E R I E N C E

In this section, attention will be given to: (1) The general procedures which are followed in stereo-photogrammetric digitising for cartographic purposes and the problems which arise during these. (2) A review of the progress which has been made in this field, including some specific examples of photogrammetric digitising for cartographic purposes.

General procedures and problems In general, most of the procedures used for setting-up and plotting stereomodels in normal stereo-compilation work may be followed when digitising for cartographic purposes, which certainly is an advantage in that only a limited re-training of personnel is necessary when introducing the method. However, because of the fact that it is, as yet, a novel method, some time has been devoted to outlining those procedures which are new and the problems which result from them. They will certainly be familiar enough to those who are currently implementing photogrammetric digitising.

Stereo-compilation. (a) Recording. After completion of the normal relative and absolute orientation, it is usual to measure and record the model co-ordinates of the control points, so that the parameters can readily be established to allow later transformation of all the recorded data to terrain co-ordinates. Next, the actual plotting mode must be decided and entered via switches on the panel of the electronics console. Three are normally available. In the point mode o] recording, individual points are recorded, e.g., at the corners of buildings, fences, and other straight line detail. Often the use of the pencil down or up positions of the operator's foot-control indicates whether the beginning or end points are being measured and a suitable code is then recorded with the co-ordinate values defining that point. The use of this method has the great advantage of saving storage space both on the output tape and later in the computer, while it is fairly easy to regenerate the complete line when plotting the final map on an automatic co-ordinatograph. The method may be used also to define curved lines, but much depends on the operator's judgement of the density of measurement, so that one of the other two modes are normally employed. These are both incremental in character, the one based on the recording of data at a pre-determined increment ol time; the other on recording after the change in distance along any one or any

PHOTOGRAMMETRIC DIGITISING FOR CARTOGRAPHIC AIMS

163

combination of the x, y and z axes relative to the previous readings exceeds some pre-set value. The time or distance increment can be selected and set on the control panel. With the former, more readings are being recorded at the slowest rate of plotting, which takes place along sinuous linear features, e.g., coastlines or contours, so that the density of data corresponds to the complexity of the feature. The method of recording distance increments ensures a more even coverage of data, if it is appropriate; also the unit will not continue to record if the operator stops to inspect a feature, as it would if a time increment is the basis for recording. (b) Feature codes. As well as the selection of the mode of recording, it is necessary to assign headers or feature identification codes so that identification or selection of the recorded data may be made for the purposes of processing, editing or plotting later. This normally involves setting the appropriate switches on the console or entry via a keyboard (e.g., for contour values). An alternative approach, which is suited to the use of the mat-type of digitiser used on stereoplotters equipped with a flat-bed plotting/tracking surface or on orthophotographs, is to use a "menu" of feature codes. With this, a set of boxes corresponding to all the classes of detail being plotted is marked out on the edge of the area being digitised. If a new type of feature is to be plotted, the operator simply places the measuring device or cursor over the box and records the co-ordinate value, which will be recognised later in the computer as the appropriate header. (c) Visual record. During plotting or compilation, there is a need for a visual record of what has been digitised. The existing co-ordinatograph or plotting table can be utilised for the purpose, especially if the plot can be seen readily by the operator: the use of the pencil-up or down mechanism to signal the beginning or end of a feature links with this well. However, there is not always a need for a precise plot in this work and, as discussed in the hardware section of this paper, there is a real need for some less accurate form of display. (d) During plotting, the normal methods of stereo-compilation apply, but perhaps a more systematic approach is necessary. For example, in normal nondigitising plotting work, a portion of a contour may be plotted first from one direction and then completed from another. Obviously, it is better from the data handling point of view if plotting of a single linear feature begins at one side of the model and goes continuously across to the other, to give fewer problems with the merging of data later. When plotting at large scales, it is often found that many features, e.g., all the buildings in a housing development, are identical. With these, it may only be necessary to plot one house at the beginning; the remaining houses are indicated by plotting only the front wall of each, together with the appropriate code, the missing sides being generated later by the computer. This procedure saves on storage space and it can also result in a considerable saving in time during the digitising process, which, together with the lack of fair drawing or touching up of the plot, may more than offset the time lost in entering headers. (e) The degree of completeness of the plot is another point of discussion. Depending on the type of terrain, scale, etc., often no more than 80% of the

164

c,. ~'ET~IV~

detail contained in the final map is generated in the stereo-plotting machine. This is then checked and completed by a field topographer via a post-plotting field completion procedure. However, the Topografische Dienst (Netherlands), N.G.O. (Norway), and other European topographic mapping agencies have shown that a comprehensive pre-plotting field completion on enlarged photographs or on overlays can greatly increase the proportion of detail plotted in the stereo-plotting machine. In a digitising context, this procedure has much to recommend it, since it would keep the post-plotting additions to a minimum, which is of considerable advantage when facing up to the problems of merging and editing data from different sources. Preliminary editing. Just as a graphical plot needs editing and touching up by the operator or his draughtsman assistant before being handed over to the cartographer, so does the digitised data. This may be distinguished as a preliminary editing as compared with the major cartographic editing which is undertaken before the final plot. Just as wrongly plotted and duplicated lines need deletion, so does the corresponding digital data. Rectangular buildings need squaring up. Redundant data, e.g., co-ordinates created while dwelling on a point in time mode, must be eliminated. If there are jumps or spikes resulting from operator mishandling, these need to be smoothed out. These jumps may even be of such an order that too high a speed of movement was prgduced for proper recordh:g of the data. These mistakes can be put right later in a computer, but the difficulty is that to make some of the changes, it is necessary to re-set the model in the stereo-plotter. Therefore, a strong case can be made out for the on-line connection of a small dedicated computer not only to act as an output device for several stereoplotters (as discussed under the heading "Recording and storage devices"), but for these controlling and preliminary editing functions. Programs can be written which execute these in real-time, if a suitable computer is available, but in any case online. An audible or visual warning can be given if a mistake, duplication or similar event occurs, and the correction applied and the erroneous record erased while the model is still in the stereo-plotter. Still other operations can be assisted by such a computer link, e.g., the numerical and absolute orientation of the model; and the direct transformation of model to terrain co-ordinates (which may be worthwhile in highway engineering work). The storage of the digitised co-ordinate information in the on-line computer memory or output devices also makes it readily available for display, either continuously or on demand for checking purposes, e.g., via a storage tube. The problem of on-line or off-line computing for control and editing purposes is a vexed one and will need thorough research over the next few years, not least into the cost-effectiveness of the respective solutions. Recently, Masry (1972) has discussed the on-line possibility in considerable detail, including in his analysis not only the on-line attachment of a computer

PHOTOGRAMMETRIC DIGITISING FOR CARTOGRAPHIC AIMS

165

to an analogue stereo-plotter (as discussed above), but the still more inter-active possibilities offered by analytical plotters with their real-time computers and feedback control operations. He shows how the real-time operating programs (i.e., the machine's software) can be modified to execute the different plotting modes (point, incremental, etc.), which are, usually, hardware-controlled in most current output units. Also the machine's plotting table can be used to plot corrections during editing operations. Some of the additions that can be made would add considerably to the real-time operating program of the AP-C type of machine, so affecting its performance, but Masry considers that recently designed, small, fast computers could be used which would overcome this problem.

Data processing. Very many possibilities exist for processing: which will be necessary depends on the form of the digitised data, the objectives it will be used to achieve, the specifications which have to be met, the computer power and software available, and many other factors. Some of the preliminary editing can be done on-line as discussed, but there are other larger tasks which will be done in a larger or more powerful computer in batch mode. For example, there is the need to merge the overlapping data obtained along the edges of adjacent stereo-models into a single unambiguous record, a task which may need some degree of manual editing via a graphical output device. Furthermore, it is necessary to add nonphotogrammetric information to the digitised record or to update it from later photography or other sources. For these tasks, a mechanical digitiser/plotter attached to the computer such as the Concord C.D.P. Cartographic Digitising Plotter, or the d-Mac Cadmac device may be useful. These devices allow the generation of hard copy and, after location and registration of the plotted data relative to the digitiser/plotter's co-ordinate system, it is possible to delete and add digitised material quite readily via back-track routines. The other obvious device for these tasks is the Cathode Ray Tube equipped with a light pen or linked through the computer to an x, y digitiser, which overcomes the defect of the low speed of plotting of the mechanical plotter, but at the expense of resolution. Again, Masry (1972) has discussed and compared some of the possibilities in his recent paper. Other problems which require study are the compaction and regeneration of data. Obvious methods of data compaction are the recording of only the beginning and end points of straight line detail, the description of a regular curve by a few individual points, the use of standard house patterns, etc. These are less applicable to the very irregular and sinuous curves of coastlines and contours where, however efficient the regenerative algorithm, there is still the need for a considerable number of points. Compaction of these may still take place by specifying the starting- and end-points and using only incremental values from those for intermediate positions. The Ordnance Survey have extended this method to long lines specifying, after transformation of the co-ordinates, the successive cutting points of the line on the National Grid and filing only incremental values between these.

166

G. PETRIE

Regeneration of the compacted data is a most difficult problem (Boyle, 1970). While the data may be held in the store in lineal form for compactness, it may need restructuring into a raster format for output in a sophisticated, high quality drum plotter such as those constructed by I.B.M, for the U.S. Army Engineer Topographic Laboratories and by C.B.S. Laboratories for the U.S.A.F. Rome Air Development Centre. Furthermore, whatever the plotting device, not only is there the question of completeness and of accuracy in position and shape of the plotted line, but there is the need for a smooth-looking final product, for the eye is extremely sensitive to lack of smoothness. A very considerable effort to ensure this is being made by several agencies and institutes. Very many more problems could be raised and discussed (in several papers) particularly with regard to generalisation, name-placement, etc., but these topics are getting outside the special competence of the photogrammetrist into the realm of the specialist cartographer and so have been omitted from this particular discussion. A further lengthy discussion of the requirements for and the characteristics of different types of automatic co-ordinatograph or plotter for cartographic purposes could also be attempted, but again this has deliberately been omitted in this paper, both on grounds of the length needed and because it can perhaps be considered to be essentially a cartographic device.

Progress and examples of photogrammetric digitising for cartographic purposes Since the whole field is still, to a considerable extent, experimental, it is only possible to give a few examples of what has been achieved to date, for, although a large number of agencies are working in this area, only a few feel that they have reached the point where they have come to any definite conclusions or that they have much in the way of final products which can be exhibited. However, it is expected or hoped that quite a number of examples or experiences will be communicated soon,

Mapping for technical (engineering) projects.

At the present time, one can say that developments using photogrammetric digitising, data processing and automatic plotting have gone furthest in this field, and two or three organisations have reached the initial production stage with their systems. The probable reason for this is that there are fewer problems of generalisation than in the topographic field where a whole set of multi-colour maps at a series of different scales, each with a different specification, may have to be derived from the initial di_gitised compilation. Also, in many cases, the final product for a technical project is often a single-colour transparency. In the engineering project field, one of the main drives towards adoption of the techniques being discussed is to increase the productivity of highly paid draughtsmen by eliminating the large amount of routine mechanical work involved in drawing, scribing, etc. and allowing them to concentrate on those tasks which need human decision-making or manual skill. The increased speed of output is a vital point when planning and executing technical projeets.

PHOTOGRAMMETRIC DIGITISING FOR CARTOGRAPHIC AIMS

167

(a) In the Ontario Department of Highways, Canada, four Zeiss Oberkochen Planimats equipped with Instronics Gradicon and Wang digitisers and magnetic tape units have been used in production on a variety of projects. Normal relative and absolute orientation procedures are used, and the stereo-compilation t • proceeds in the usual way. For recording data, either the point or time mode is used, the latter having a selectable rate of recording for curved lines, contours, etc. Feature coding is by selectable thumbwheel switches on the digitiser electronics or by keyboard entry: names, etc. are also entered by keyboard. The raw digital data is cleaned up, verified and edited off-line on a Hewlett Packard 2116B computer with 16K storage. The same computer is used as controller for the Gerber 32 co-ordinatograph equipped with an optical head for exposure of photosensitive material for the final map. In addition to numerous large-scale contoured plans for engineering work, medium-scale topographic sheets have been produced. An outstanding example is the 1 : 25,000 scale topographic map of Peterborough, Ontario, and its environs. The initial colour photography was taken at 1/50,000 scale with a Zeiss Oberkochen RMK f = 85 mm super wide-angle camera. The four stereo-models required were compiled and recorded digitally on magnetic tape in a total time of 190 hours. Final plotting of the sheet on the Gerber co-ordinatograph occupied 26 hours; a further 32 hours were used in manually preparing the area masks using peel coat material. (b) The Meetkundige Dienst (Surveying Office) of the Rijkswaterstaat (State Service for Roads and Waterways), Netherlands, has also been utilising a digitised procedure in production for two or three },ears (Van den Hout, 1970). Digitising is undertaken on Wild A8 stereo-plotters equipped with EK-5 and EK-8 co-ordinate registration equipment with Wild SL-15 and Facit 4070 paper tape punches. Till recently, continuous recording on a time or distance base has not been employed, a curve being defined by a series of individual points selected by the operator. Compared with any other agency's procedure for mapping, this office is distinct in that it does not execute an absolute orientation in the stereoplotter. The planimetric detail and the aerial triangulation and ground control positions are, in fact, all measured during a single operation. Later, when the strip or block is completed, the aerial triangulation is computed, and processing of the remaining digital data, including transformation of the detail to terrain system, can begin. Plotting has, till now, mainly been carried out on Calcomp drum plotters. By this procedure, only a single set-up of the model is necessary and work can begin ahead of the ground control operations. On the other hand, processing and plotting is not possible till the whole strip or block is completed, which is a disadvantage with a large block of photography. The maps produced by Rijkswaterstaat for the mainly flat Dutch countryside are comprised of planimetric detail and a series of individual spot heights, which are readily transformed to the ground system. So the complexities and difficulties, which would be en-

168

G. PETRIE

countered with the transformation of measured contour lines of unknown scale and orientation in rough terrain, are avoided, in particular the problem of fitting them to streams, lakes and planimetric detail.

Topographic mapping. In this area, a great deal of work has been undertaken in a variety of countries--particularly in the U.K., Canada, U.S.A. and Sweden-and a number of experimental sheets have appeared from each. As already discussed, there are usually a variety of sources of data for the production of topographic maps, e.g., existing larger scale maps, as well as a new stereo-compilation. Mostly, the experimental sheets produced till now have used existing maps as source material, with x, y cartographic digitisers providing the digitised input for the process. However, some have used photogrammetric material from the outset and these will be discussed here. (a) The U.S. Army Engineer Topographic Laboratories have produced an experimental 1 : 1 2 , 5 0 0 scale multi-colour topographic sheet of the Fort Belvoir, Virginia, area adjacent to the Laboratories. The stereo-plotter used was a Kelsh Plotter equipped with Dell Foster digitising equipment (resolution: 0.005 inch = 0.125 mm) and a magnetic tape recorder (Babcock, 1970). The normal procedures for stereo-compilation were used, with the addition of new headers and codes whenever a new class of data was measured. Their insertion adds about 25% to the normal time for stereo-compilation, but Babcock makes the point that a good deal of inking or re-tracing of the plotted manuscript is normally carried out after compilation which is eliminated when digitising, a saving which more than counterbalances the 25% time increase required for coding. Furthermore, a definite saving in time occurs using the digitising method where a large number of standard symbols have to be plotted. For example, 350 standard building symbols in one stereo-model required 130 min to compile by normal methods, but only 40 min using the digitised method. For the Fort Belvoir map, some seventeen models were digitised, taking an average of only 6 h each, since not too much detail was required. The data was compiled, scaled and joined in an XDS-930 computer which also generated all the required symbols. The final colour-separated plots were produced on a Calcomp 702 high-speed, flat-bed plotter, a total of only 31/2 h being required for the computer processing, editing and plotting of the 1 : 12,500 scale map. The Calcomp plotter was used simply to prove the system and data; a Gerber coordinatograph with optical exposure head is now available in the Laboratories to produce higher quality output. (b) The Cartographic Unit of the U.S. Soil Conservation Service, Hyattsville, Maryland, has produced a series of experimental sheets using digitised data (Johnson, 1972). Digitising is carried out using Dell Foster equipment, which can be mounted either on a Wild B8 Aviograph or over a light-table on which can be mounted existing maps, orthophotographs, etc. Digitising starts with line work: three classes of roads, four types of drainage, etc., are identified, measured

PHOTOGRAMMETRIC DIGITISING FOR CARTOGRAPHIC AIMS

169

and recorded. Next, the soil information is plotted, starting with line boundaries annotated on the photographs and then proceeding to the alphanumeric information required for names, symbols, etc. After processing, plotting is carried out on a Gerber co-ordinatograph. Johnson mentions that the above method has been used for the production of eleven sheets in Crawford County, Kansas. (c) The Surveys and Mapping Branch of the Department of Energy, Mines and Resources, Canada, has also progressed to a pilot production stage. Harris (1971) argues the case for a wholly on-line system with digitisers and co-ordinatograph linked directly to a medium-size computer. The Surveys and Mapping Branch is in the process of implementing this approach, but in the initial stages, the photogrammetric component of the system, a Wild B8 Aviograph equipped with a digitising system constructed in-house, is attached to a PDP-11 minicomputer with an 8K memory, equipped with a magnetic tape recorder. The PDP-11 is capable of scanning the output from a variety of digitising devices, so forming a complete off-line digitising system, but it is planned to connect it eventually to the main computer. Till now, most digitising in this Canadian Federal mapping agency has been carried out with d-Mac, Instronics Gradicon and Bendix Datagrid cartographic digitisers which have been placed on-line to the Survey's PDP-10 computer with 32K core memory. The draughting table used is a Kongsberg Kingmatic equipped with a photo-exposure head. At present, this utilises a mini-computer as a Director, but presumably the whole unit will be interfaced with the PDP-10. Experimental topographic sheets have been produced from this system, e.g., a sheet covering the area around Press and Sturgeon Lakes, northern Ontario. (d) The Ordnance Survey, U.K., has, like some of the other organisations above, been using d-Mac and Ferranti Freescan cartographic digitisers as the main source of its digitised data, but it has also equipped a Wild A8 Autograph with a d-Mac digitising system with alternative higher-speed paper tape and magnetic tape output units for its experimental work (Sowton, 1971). The Ordnance Survey has conducted an experiment, in co-operation with the Experimental Cartography Unit, in which material has been digitised at 1 : 2,500 scale on a d-Mac table for 100 kin: of the Bideford area in southwest England. This has been used to produce much of the detail for the family of medium scale (1 : 10,000 and 1 : 25,000 scale) maps which are derived from it, and a 1 : 25,000 scale map plotted using an AEG Geagraph automatic co-ordinatograph. A detailed discussion of the problems encountered in this work is given in Irwin (1971). However, most of the Ordnance Survey's effort has been applied to the large scale 1 : 1,250 and 1 : 2,500 planimetric series which cover all urban areas in the United Kingdom (Gardiner-Hill, 1971). These maps are of great interest to local planning authorities and public utility agencies, who do not necessarily require all the planimetric information provided on the standard published sheets and need to add a lot of specialised information of their own, which needs frequent updating. So a digitised system of recording topographic information can be

170

6. PETRIE

utilised not only by the mapping agency, but can readily be used (and sold) to a variety of users. A variety of different 1 : 1,250 scale maps have been plotted for different areas in London, using a Calcomp 24-vector drum plotter for rough products and a Ferranti Master Plotter, equipped with an optical head, for maps plotted and drawn to full production specification. One area where a considerable development has taken place since the last Congress is in the preparation of programs which will allow contours to be interpolated and plotted from a series of individual spot heights. These may be the individual points measured during digital terrain model work, but, more usually, they are measured and recorded on paper or magnetic tape during the profiling carried out for orthophotograph production. Several different programs exist (e.g., Konecny and Refoy, 1968), but the most highly developed so far are the CONPLOT programs developed for the U.S. A r m y TOPOCOM for use with profiles measured in the UNAMACE machines. The major difficulty with such programs is in fitting the interpolated contours to the planimetric features, especially to drainage, roads ao_d railv/ays. Whether it would be better to measure the contours in the stereo-plotter and then interpolate height profiles from these for the control of orthophotograph production off-line is an obvious area for research at the moment. Contours from digitising profile in[orrnation.

SUMMARY As can be seen from the above account, the application of digitising of photogrammetric measurements for cartographic applications is only in its infancy and mostly experimental in character till now. However, it is a field in which a good deal of progress has been made in the last four years, and one to which a great deal of attention will be paid in future. Its eventual widespread adoption can hardly be doubted. REFERENCES Anonymous, 1970. Proceedings of the Symposium on Map and Chart Digitising. U.S.G.S. Computer Contrib., 5:81 pp. Balzcock, H. C., 1970. Evaluation of a stereo-compilation digitiser. Proc. Ann. Meeting Am. Congr. Surv., Mapping; 30th: 338-347. Bailey, K. V., Forrest, R. B. and Hanaway, D. P., 1969. Bendix Datagrid Digitiser. (Paper presented at A.S.P.-A.C.S.M. Cony., March 1969:8 pp.). Boyle, A. R., 1970. The quantised line. Cartogr. J., 7(2):91-99. Forrest, R. B., 1972. A digital portable line-drawing rectifier. Bendix Tech. J., 5(1):67-68. Gardiner-Hill, R. C., 1971. Automated cartography in the Ordnance Survey. Proc. Conf. Commonw. Surv. Officers, Pap., E3; 15 pp. Harris, L. J., 1971. Automated cartography in federal mapping in Canada. Proc. Conf. Commonw. Surv. Of/icers, Pap., E2; 16 pp. Hattaway, D. P., 1971. An all-electronic two-axis digitiser. Bendix Tech. J., 4(1):53-54. Howard, S. M., 1968. A cartographic data bank for Ordnance Survey maps. Cartogr. J, 5(1):48-53.

PHOTOGRAMMETRIC DIGITISING FOR CARTOGRAPHIC AIMS

171

Hughes, T. C., Shope, A. R. and Baxter, F. S., 1971. U.S.G.S. automatic orthophoto system. Photogramm. Eng., 27(10):1055-1062. Irwin, M. S. G., 1971. Developments in automated cartography at the Ordnance Survey. Cartogr. J., 8(2):133-137. Johnson, C. G., 1972. The proposed advanced mapping system for the Soil Conservation Service. Proc. Ann. Meeting A.S.P., 38th: 152-167. Konecny, G. and Refoy, D. H., 1968. Maps from digitised Stereomat data. Photogramm. Eng., 34(1) :83-90. Masry, S. E., 1972. Real-time digitising and editing from photogrammetric instruments. U.N.B. Tech. Rep., 12:55 pp. Petrie, G., 1971. Photogrammetric digitising: input for data processing. I.T.C. Publ., Ser., A, 50:59-92. Sowton, M., 1971. Automation in cartography at the Ordnance Survey using digital output from a plotting machine. Bildmessung Lu/tbildwesen, 39(1):41-42. Uhrig, H., 1970. Untersuchungen~ zum Datenumfang und Speicherbedorf sowie zur automationsgerechten Gestaltung der Zeichen fiir die Topographische ~bersichtskarte 1:200,000. Nachr. Karten-Vermessungswesen, Set'. 1, 47:61-73. Van den Hout, C. M. A., 1970. Digitale Grundrisskartierung in grossen Masstaben. Bildrnessung Lu/tbildwesen, 38(1):85-90.