- Email: [email protected]
Digital images in the map revision process
PHOTOGRAMMETRY & REMOTESENSING ELSEVIER
ISPRS Journal of Photogrammetry& RemoteSensing51 (1996) 188-195
Digital images in the map revision process P.R.T. N e w b y GEO-UK Ltd, 9 Merrytree Close, West Wellow, Romsey, Hants S051 6RB. UK
Accepted 17 January 1996
Abstract Progress towards the adoption of digital (or softcopy) photogrammetric techniques for database and map revision is reviewed. Particular attention is given to the Ordnance Survey of Great Britain, the author's former employer, where digital processes are under investigation but have not yet been introduced for routine production. Developments which may lead to increasing automation of database update processes appear promising, but because of the cost and practical problems associated with managing as well as updating large digital databases, caution is advised when considering the transition to softcopy photogrammetry for revision tasks.
1. Introduction It is clear from the current literature that photogrammetry is in the throes of a revolution (Welch, 1992; Heipke, 1995). The means now exist to leave behind the silver halide graphic images on which we have relied since the beginning of photogrammetry, and to manipulate instead a digital model of the world. Airborne sensors may eventually follow satellite-borne sensors into the digital data era. However, the state of the art of photogrammetry (as it affects the revision or updating of maps and of digital topographic databases) is still to make aerial photographs on polyester-based silver halide emulsions using cameras which would be broadly recognisable to the photographers of fifty years ago. The revolution so far only affects the subsequent stages of processing. As in all revolutions, forces of reaction also are at work, based in this case on sound economic and practical grounds. Thus, in this paper, no assumptions are made about the outcome. Instead, the state of development of map and database
revision towards the end of the 1992 to 1996 session of International Society for Photogrammetry and Remote Sensing (ISPRS) is reviewed, and some examples of promising research activities are cited,
2. The transition from analogue to digital photogrammetry Developments in computer science over the past thirty-five years have provided the driving force for new surveying, mapping, and photogrammetric techniques and equipment. The first productionised analytical aerial triangulation system which used electronic recording stereocomparators and "highspeed' computers largely mimicked an existing system based on the use of manual recording stereocomparators and hand calculators while greatly extending its possibilities (Arthur, 1959; Proctor, 1962). The subsequent astonishing developments in processing speeds and on-line storage, together with equally remarkable reductions in cost, have brought us through digital (or numerical) cartography, computer-assisted
0924-2716/96/$15.00 © 1996ElsevierScienceB.V. All rights reserved. Pll S0924-27 16(96)0001 2-3
PR.T. Newby /ISPRS Journal of Photogrammetry & Remote Sensing 51 (1996) 188-195
digital data capture using analogue stereo-plotters equipped with encoders, and later, analytical stereoplotters in which optical and mechanical parts were minimised and the photogrammetric restitution was carried out in real time by computational means. As numerical cartography gradually evolved into today's topographic databases and geographical information systems (GIS), graphical map output from photogrammetric plotters became less relevant. Moreover, those who already possessed topographic databases soon recognised that their concern in the future would be maintaining an up-to-date database. Thus, by the end of the 1980s the photogrammetric plotter had become a photogrammetric workstation at which a human operator, assisted by one or more computers, could capture or edit three-dimensional geometric, topological and semantic data about the world portrayed in collections of high-quality, twodimensional graphic images and recorded in digital databases (Bonjour and Newby, 1990). Meanwhile our colleagues developing the newly emerging discipline of remote sensing were employing digital images transmitted from satellites. Developments in computer science parallel to those exploited by surveyors and mappers allowed them to handle such images. Image processing and computer vision developed independently of photogrammetry for the simple reason that photogrammetrists retained the major advantage of working with graphic images of very high geometric and radiometric quality. Only very recently have developments in computer graphics made digital images worthy of the attention of photogrammetrists. Now there is a most welcome convergence between photogrammetry and remote sensing, image processing and computer vision. It is clear that photogrammetry can benefit from existing expertise in digital image processing, while contributing traditional strengths in the rigorous understanding and manipulation of the geometry of image formation. Digital (or softcopy) photogrammerry has now developed to the point where numerous vendors offer viable systems, and civilian mapping organisations in some of the world's more advanced countries have at least tried some aspects of digital photogrammetry, even if they have not yet adopted them whole-heartedly (Colomer and Colomina, 1994; Heipke, 1995). For map revision there are obvious possibilities
189
of using either the digital orthophoto as a data source for monocular data capture, or a digital photogrammetric workstation (DPWS) with stereoscopic viewing as a three-dimensional edit station. To this author's knowledge, however, there are as yet no organisations using softcopy photogrammetry as a routine production tool for map or database revision. Looking further ahead, researchers in several countries have made progress in automating change detection and hence feature extraction from digital images. Although there are some examples of this work, it is likely to be a long time before such techniques leave the research laboratory and enter the daily routine of those who earn their living by maintaining topographic databases. In the real world of routine production, all technical processes stand or fall by their cost effectiveness. This is especially the case with entirely new developments such as softcopy photogrammetry that involve high initial investments for research and development, capital equipment, staff training and (most importantly for database revision) integration with existing systems. To justify changes, cost benefits must be demonstrable. The following section will illustrate how this has worked in practice in a leading national mapping organisation. 3. Current map revision practice, with special reference to Great Britain
The large scale digital database in Great Britain is now complete. Approximately 60,000 sheets originally mapped at 1:1250 scale, some 160,000 at 1:2500 scale and almost 2000 at 1:10000 scale have been converted to about 220,000 digital files in which distinctions between old style map scales are becoming increasingly blurred. Moreover, all change on the ground of major importance to the map user is surveyed and the database is updated accordingly within six months of the change occurring. This represents a significant evolution of the practice of the Ordnance Survey of Great Britain (OS) in response to one of the world's most sophisticated communities of civilian customers for topographic information. Much has been written about OS revision practice in recent years (for example, Newby and Proctor, 1990; Newby, 1990a; Newby, 1994; Vincent and Logan, 1995). This section will provide a summary of
190
P.R.T. Newby/ISPRS Journal of Photogrammet,3, & Remote Sensing 51 (1996) 188-195
developments in both practice and policy, the result of technical research and development, consultation with customers, and the relentless pressure from government to reduce costs, increase efficiency, and involve private enterprise in its activities. For many years OS boasted of its system of continuous revision of major change, but in practice this nevertheless allowed substantial backlogs to build up, and minor change could take as much as forty years to be captured. The customers for OS digital data, or for the graphics derived from that data, are now more demanding, hence the emergence of the new system mentioned above, termed 'low threshold revision', by which major change reaches the database within six months of its occurrence on the ground. Reaching the present position involved a hectic catching-up exercise, most cost-effectively carried out using analytical photogrammetric workstations with superimposition of the existing digital data. In order to meet tight deadlines, revisions in many areas were done by field surveyors. What should in future constitute 'major change' has been the subject of lively debate as well as consultation with customers; however, most major change in future will surely be surveyed in the field because by definition 'low threshold' implies small increments. The exception will be the survey of major development projects of short duration, where timely photography will enable economic photogrammetric revision within the six month time frame. It is also possible that aerial photographs will be used as an aid to detection of urban change. Meanwhile minor change will move from being a continuous (but generally infrequent) process to a cyclic process at 5 year intervals (10 years in mountain and moorland areas), beginning with the clearance of outstanding backlogs. Thus, the new revision strategy which has emerged over the past four years can be summarised as follows: improve urban continuous revision, introduce cyclic revision for rural and mountain areas, and extend continuous revision of major change to rural areas. The cyclic revision of rural areas for minor change will be largely a photogrammetric process, both with regard to the detection and the survey of change. Change on the ground affects some 2% of detail in the database every year. Detecting that change, and verifying the absence of change from the
remaining 98%, both form part of the revision process, both incur significant costs, and neither can be neglected. With 160,000 map files for rural (1:2500 scale) areas, a five year cycle implies the revision of over 30,000 files per year. At any one time, hundreds or even thousands of files may be in work. Just managing these data is a demanding task. Data transfers, backups and validations require discipline, quite apart from the process of update itself, and the topological structure verification that must follow every update operation. This, of course, is for vector data. As soon as softcopy images are introduced, the volume of data will increase by up to three orders of magnitude as vector map files of typically well under a megabyte (Mb) are backed by images of the order of 100 Mb each. The processes used to survey change have been comprehensively documented. Newby (1990b) reported trials which led to the acceptance of the analytical 'plotter' with superimposition as the most cost effective means of updating OS large scale maps for a wide range of densities of change. However, all of the methods tested were found to be workable and were actually used over a protracted period for historic reasons associated with the legacy of earlier generations of equipment. These methods ranged from pure graphic plotting using analogue plotters, with subsequent digital data capture at a conventional digitising table, through on-line editing at a workstation linked directly to an upgraded analogue or an analytical plotter, to the current method of choice with the analytical plotter and superimposition. Here the operator's view through the eyepieces includes the three-dimensional landscape as portrayed on silver halide diapositives exposed using a modem camera with forward motion compensation, the old and new database (map) detail, and the menu for the editing process using LaserScan's LITES2 software. This is a cost-effective process which points towards a major advantage of softcopy systems where the superimposition of digital image and line-map detail from the existing database will come as part of the system at no extra cost. Although each successive development has smoothed the process of on-line updating, in practice a final edit is always done at a comparatively inexpensive two-dimensional graphics workstation away from the photogrammetric workstation itself.
I~R.Z Newbv /lSPRS Journal of Photogrammetry & Remote Sensing 51 (1996) 188-195
What now of the future? At the time of the last OS procurement of photogrammetric revision equipment, DPWS did already exist in the marketplace. It was considered premature to adopt them for database revision then, because their integration with existing data management practices would have delayed their effective contribution to clearing revision backlogs. Now, two years on, any organisation purchasing photogrammetric equipment for database revision should at least consider the numerous available DPWS. Meanwhile, OS has been conducting extensive developments and trials of digital monoplotting processes (Vincent and Logan, 1995). These use a digital (softcopy) aerial photograph obtained by scanning a conventional negative or diapositive, together with a digital terrain model (DTM) which enables the correct geometry to be extracted without the use of stereoscopic viewing. In principle either the raw digital image or a digital orthophoto may be used. The software for superimposing the existing data and updating the database with newlysurveyed change will obviously be simpler in the latter case. OS have disclosed that both Intergraph and LaserScan systems (the latter using DEC Alpha workstations) have been in use in the trials, but results have not been published. However, a recent public invitation to tender for a new photogrammettic scanner allows us to speculate that their digital photogrammetric trials, which date back over six years, will soon bear fruit in the production arena (Newby, 1990c). Since a great deal of OS revision work is done by field surveyors it is also worth mentioning the processes used. The amount of ground control depends on the extent of the area being revised and the nature of existing surrounding detail. If new control is required, it will be supplied by a suitable combination of total station traversing and Global Positioning System (GPS) surveys. Some new map detail may also be collected by total stations, but complete map data capture by electronic recording at the total station has not found as much favour among OS field revisers as among engineering project surveyors. Instead, a graphical survey method peculiar to OS is used to complete the new map detail, with or without the assistance of rectified enlargements of aerial photographs, followed by database update on a graphics workstation with digitising tablet in the field office.
191
This leads us to an exciting development in field surveying - at the opposite end of the scale from developments in GPS - the PRISM (Portable Revision and Integrated Survey Module) project based on a pen computer. This has the potential to become a complete, portable, interactive edit station for map data, replacing the graphic processes used hitherto. After several years of development PRISM is now in intensive production trials with some eighty units in use (Greenway, 1994; Vincent and Logan, 1995). If this proves to be a cost effective tool it will shift the economic balance between field and photogrammetric processes. Such a device could eventually incorporate data from total stations or GPS, or indeed carry a digital image backdrop as well as line map data, thereby placing a DPWS in the hands of the field surveyor and blurring historic distinctions between field and photogrammetric surveys! The government of Great Britain is increasing the role of the private sector in its map revision activities. Revision is now being done by a mixture of in-house work and private sector contracts, the latter primarily intended to assist rural cyclic revision. Intense competition in the market place and European Community rules on public contracts have led to contract awards to private sector companies not only from the United Kingdom (including Northern Ireland) but also from Denmark. It remains to be seen how the private sector, especially from outside Great Britain, copes with the peculiar disciplines of national database update. There is clearly a strong intent to make the process a success, and an enlightened application of quality systems and standards such as BS EN ISO 9000 and ISO 2859 in its support. In the absence of published information we can only speculate on the revision techniques. However, although product rather than method specifications are being used, it seems likely that contractors will be playing safe with well-trusted classical procedures rather than pushing back the frontiers with new developments.
4. Aerial images and their conversion to digital images It is not feasible to enter digital photogrammetry half-heartedly. To justify the investment in a scanner to convert photographic hard copy images into soft-
192
P.R.T. Newby / ISPRS Journal of PhotogrammetO' & Remote Sensing 51 (1996) 188-195
copy it is necessary to plan to make use of its full capacity. Unless reliance is to be placed on an agency service for photo scanning, a substantial investment in both a scanner and numerous DPWS at the same time is required. There has been extensive speculation on whether it is necessary to spend the large sums required for a top quality scanner designed expressly for photogrammetry, or whether much lower cost systems designed for less demanding tasks, or at least using off the shelf components, can give acceptable results. Certainly, it is now recognised that not all scanners are the same, that even very expensive scanners may not provide perfect results, and that the complete digital photogrammetric flowline will depend on this single crucial component (Bill, 1994). The importance of this issue has led to the formation of a joint European Organization for Experimental Photogrammetric Research (OEEPE)-ISPRS working group on the analysis of photo-scanners. Some key factors in the performance of scanners were propounded at an early meeting of that working group (K61bl et al., 1994). These covered: (1) the technical requirements of geometry, resolution, image noise, dynamic range and so on; (2) comfort and convenience aspects, such as the level of automation of the process and the use of original roll film negatives; and (3) engineering points involving the satisfactory solution of classical problems of aerial imagery such as optical distortion, image flatness, vibration and the avoidance of artefacts. Above all, the results of scanning must aim to be as good and cheap as a diapositive! This makes it quite hard for digital photogrammerry to compete. Photogrammetric operators have only just got used to the improvements which followed innovations in aerial photography such as new lens designs, forward motion compensation (FMC) and higher resolution emulsions - all of which have led photogrammetrists to expect sharper, better images than ever before. Moreover, for the surprisingly demanding task of detecting and surveying change while minimising the need for subsequent field completion, exploiting the improved geometric accuracy of modem photographs and measuring systems by reducing the scale of photography has not proved a sensible course. Instead, ease of interpretation has become the dominant requirement. Throwing all of this away in exchange for the privilege of viewing
lower resolution pixel-based images on a computer screen requires an act of faith or a dramatic cost advantage. Nevertheless, Stirling (1995) has reported that users of digital systems do adjust to the comparatively fuzzy screen images and that the loss of information is not severe. This introduces the question of the trade-off between pixel resolution and data volume which affects the cost of capture, storage and transmission of data. The standard was set by the Zeiss/Intergraph PhotoScan, which allows a smallest pixel size of 7.5 /~m. This implies about 600 Mb of data per aerial image. Much discussion of the minimum practical resolution has ensued, but no firm consensus has yet emerged. However it seems likely that most users will eventually settle for pixels in the 15 to 25 # m range, or some 100 Mb per image (Welch, 1993). Whether or not the next generation has difficulty with the storage volumes involved (terabytes per year for the national mapping of a medium sized country), processing speeds and transmission rates will certainly be critical, as will the disciplines of managing such quantities of data. Until very recently the idea of replacing the conventional aerial camera with a digital system in the aircraft seemed ludicrous. Developments over the past three years counsel caution in dismissing such ideas. The Institut Geographique National (IGN) (France) has been experimenting with airborne digital cameras and, through OEEPE, is seeking collaborators in order to develop the concept further. Meanwhile, in North America, systems are in use for environmental monitoring and are producing images which show some potential even for urban map revision (Monday et al., 1994). At the same time there are plans for commercial, satellite-borne, highresolution imaging systems as discussed by Konecny and Schiewe (1996) in this issue. With pixel sizes of one to two metres on the ground these could be of interest for small-scale map revision, but they will surely not replace airborne imagery for the update of large-scale databases. 5. DPWS hardware and software considerations At its most basic level the DPWS has monocular viewing only, of a single image whose geometry is corrected by a good DTM. Any monocular system
PR.T. Newby /ISPRS Journal of Photogrammet O, & Remote Sensing 51 (1996) 188-195
deliberately discards one of the most crucial benefits of photogrammetric surveying, namely stereoscopic viewing. Of course, this serious disadvantage may be outweighed by the obvious simplicity and economy of monoplotting systems. If not, it will be necessary to wait for the hand-held field computer to combine monocular photogrammetry with the surveyor's eye view at ground level before the loss of stereoscopy will be tolerated by most photogrammetrists. At the next level of present-day development, with the addition of stereoscopic viewing, the DPWS replicates the functionality of an analytical plotter. Like the old universal analogue machines, in its most advanced form it will be able to be used for all known photogrammetric processes. All DPWS will have the facility of viewing the existing vector or raster map data, together with any new c~ata captured from the current aerial image. For revision purposes the capture and edit system must integrate easily with the existing database or GIS. This integration points to a major advantage of digital photogrammetry in terms of quality management - real world geometry will always be maintained, or at least departures from it will be immediately apparent. Along the way the DPWS will be able to generate DTMs automatically and will allow them to be edited in a user-friendly way. From the DTM, orthophotos will in turn be generated automatically (Schiewe and Siebe, 1994). In order to use orthophotos of urban areas for map revision (or indeed any other purpose) it is essential that they show the tops, as well as the bases, of buildings in their correct planimetric positions. A solution to this obvious requirement, long overlooked, was described by Meister and Dan (1994) at the Commission IV Symposium held in Athens, Georgia. An up-dated report on the procedures is presented by Sinning-Meister et al. (1996) elsewhere in this issue. At least one leading vendor (Leica) now includes a similar procedure in a DPWS, although the onus remains on the user to capture the necessary three-dimensional building model manually in order to create a displacementfree orthophoto. Once the 'building lean' problem is eliminated automatically, the floodgates will open for the acceptance of orthophotos for urban mapping and database revision.
193
6. Other promised advances in map and database revision Almost all of the processes described so far have still relied on the involvement of the human operator. To contain the cost of database maintenance in a cost-conscious world, researchers are looking increasingly to automation. The excellence of the human eye and brain (or what soldiers call 'the Mark 1 eyeball') at tasks of image understanding and pattern recognition, however, is even harder to reproduce in a computer than it is to understand. Nevertheless, some progress has been made in recent years that points to the possibility of at least semi-automated processes reaching the production arena. One of the less demanding tasks in image understanding is to recognise and follow linear features such as roads. Successful demonstrations have been given using both SPOT and aerial image data (De Gunst et al., 1991: Solberg, 1992; Sakoda, 1993; Plietker, 1994; Peled, 1994). Comparison of a new image with the old network allows the latter to be updated. Peled has outlined a plausible scenario for future progress, beginning with a semi-automatic process of subtraction of old and new images followed by noise removal, to supply the human reviser with candidate areas for his attention. Next, in what he calls 'GIS-driven updating,' the above process would provide the trigger for automated recognition of new detail and its extraction in a hierarchy of themes. Finally, autonomous rule-based artificial intelligence (AI) systems may largely take over from the human operator. As in most aspects of image understanding, it seems unlikely that any one algorithmic approach will yield the required results, but that a combination of approaches, mimicking the human's intuitive combined tactics, may eventually be successful. Similar efforts have been reported in the detection and extraction of buildings, notably by McKeown et al. (1985) and Murakami and Welch (1992). Thus far, the distinction between two- and threedimensional databases has not been fully defined. A map models a three-dimensional world and, until recently, most GIS developers have preferred to treat it as two dimensional. Certainly the topological structure becomes very much more complex if the third dimension is allowed to intrude. Theoreticians have
194
P.R.T. Newby /ISPRS Journal of Photogrammetry & Remote Sensing 51 (1996) 188-195
now started to wrestle with the question of whether time ought to be treated as a fourth dimension in a GIS or whether update merely generates a succession of states of the database (which may or may not be stored for posterity). If the latter view is accepted, it makes it easier for us to consider practical matters such as change on the ground which has not yet reached the database, change in the database as a result of improved data without any corresponding change on the ground, and the requirements of users for complete new versions of the database after update or merely replacement of the updated features. As data structures become more rigorous and complex, these matters pose formidable problems of data integrity for both supplier and customer. They are necessarily addressed in practice by organisations like OS, but are also now receiving theoretical attention from researchers such as Kemppainen (1994).
7. Conclusions Digital or softcopy tools will undoubtedly be used for revision in the future but must compete effectively on cost grounds as well as being tailored to interact with existing database structures and practices, for this is the essence of revision as opposed to original survey. It is not just the storage but also the manipulation and transmission of the huge volumes of data entailed in digital images which will demand careful attention. The question of monoversus stereo-viewing is an important issue today, but in this author's opinion stereoscopy is a special advantage of photogrammetry which will not lightly be given up, at least until developments in data handling and transfer allow the field surveyor to hold a (monocular) DPWS in his hand, on site. Database revision has now become a respectable discipline within the ISPRS commission structure. Technical and institutional aspects rank equally as worthwhile subjects for discussion. There has been perceptible and promising progress towards the longterm goals of automated change detection and feature extraction but practical solutions are still a long way off. Database integrity after update is now attracting serious attention, being helped in practice by the use of the best tools and the best possible imagery as an aid to interpretation. The forthcoming Vienna Congress of ISPRS and the next session from 1996
to 2000 should provide interesting developments in this very important practical subject.
Acknowledgements The author was formerly Manager of Photogrammetry and Survey Computations at the Ordnance Survey of Great Britain. He acknowledges his debt to that organisation and to other former employers both for his photogrammetric experience and for the opportunity to begin serving ISPRS. However, he is now obliged to rely on published sources for reports on new developments. The opinions expressed in this paper, and any errors which may inadvertently have been included, are therefore entirely the responsibility of the author and have no connection with the views of the Ordnance Survey.
References Arthur, D.W.G., 1959. Recent developments in analytical aerial triangulation at the Ordnance Survey. Photogramm. Rec., 3(14): 112-124. Bill, R., 1994. Scanning of hardcopy images: achievements and limitations. In: Heipke and Mussio (Editors), ISPRS Commission, I Tutorial Notes, Acquisition, Characterization and Archiving of Digital Imagery, 62 pp. Bonjour, J.D. and Newby, ER.T., 1990. Architecture and components of photogrammetric workstations. Photogramm. Rec., 13(75): 389-405. Colomer, J.L. and Colomina, I., 1994. Digital photogrammetry at the Institut Cartogrhfic de Catalunya. Photogramm. Rec., 14(84): 943-956. Greenway, I., 1994. The use of pen computers for revision of large scale mapping. Presented to 20th FIG Congress, Melbourne, Commission 5, paper 505.4, 10 pp. De Gunst, M.E., Hart, C.S.L.A., Lemmens, M.J.EM. and Van Munster, R.J., 1991. Automatic extraction of roads from SPOT images. Proceedings of ISPRS and OEEPE joint workshop on updating digital data by photogrammetric methods, Oxford. OEEPE Official Publication 27: 131-140. Heipke, C., 1995. State-of-the-art of digital photogrammetric workstations for topographic applications. Photogramm. Eng. Remote Sensing, 61(1): 49-56. Kemppainen, H., 1994. Modelling update propagation in spatial databases. Int. Arch. Photogramm. Remote Sensing, 30(4): 625-633. K61bl, O. et al., 1994. Documents of the OEEPE-ISPRS Joint Workshop on the Analysis of Photo-Scanners, EPF-Lausanne. McKeown, D.M., Harvey, W.A. and McDermot, J., 1985. Rulebased interpretation of aerial imagery. IEEE Trans. Pattern Recog. Mach. Intell., PAMI-7: 570-585. Meister, M. and Dan, H., 1994. Processing of geographic data
P.R.T. Newby / ISPRS Journal of Photogrammetry & Remote Sensing 51 (1996) 188-195 for CAAD-supported analysis and design of urban areas. Int. Arch. Photogramm. Remote Sensing, 30(4): 433--440. Monday, H.M., Urban, J.S., Mulawa, D. and Benkelman, C.A., 1994. City of Irving utilizes high resolution multispectral imagery for NPDES compliance. Photogramm. Eng. Remote Sensing 60(4): 411-416. Murakami, H. and Welch, R., 1992. Automatic feature extraction for map revision. Int. Arch. Photogramm. Remote Sensing, 29(B4): 569-575. Newby, ER.T., 1990a. Digital map revision at the Ordnance Survey. Int. Arch. Photogramm. Remote Sensing, 28(4): 146154. Newby, ER.T., 1990b. Photogrammetric developments in the Ordnance Survey in 1990. Photogramm. Rec., 13(76): 561576. Newby, ER.T., 1990c. New initiatives in image digitising at the Ordnance Survey. Int. Arch. Photogramm. Remote Sensing, 28(3/2): 639-656. Newby, ER.T., 1994. Revision policy and practice for Great Britain's topographic database. Int. Arch. Photogramm. Remote Sensing, 30(4): 260-267. Newby, ER.T. and Proctor, D.W., 1990. Revision of large-scale maps at the Ordnance Survey. ISPRS J. Photogramm. Remote Sensing, 45: 137-151. Peled, A., 1994. Revision of digital maps and GIS databases. Int. Arch. Photogramm. Remote Sensing, 30(4): 268-272.
195
Plietker, B., 1994. Semi-automatic revision of street objects in ATKIS database DLM 25/1. Int. Arch. Photogramm. Remote Sensing, 30(4): 311-317. Proctor, D.W., 1962. Adjustment of aerial triangulation by electronic digital computers. Photogramm. Rec., 4( 19): 24-33. Sakoda, B., 1993. Computer-assisted tracing of faint roads on satellite images. Proc. ACSM/ASPRS Annual Convention, New Orleans, 2:311. Schiewe, J. and Siebe, E., 1994. Revision of cartographical databases using digital orthoimages. Int. Arch. Photogramm. Remote Sensing, 30(312): 724-731. Solberg, R., 1992. Semi-automatic revision of topographic maps from satellite imagery. Int. Arch. Photogramm. Remote Sensing, 29(134): 549-556. Stifling, D.M., 1995. Review of the use of digital workstations in the UK. Technical meeting of the Photogrammetric Society, London, 24 October 1995. (To appear in Photogrammetric Record, 1996.) Vincent, G.N. and Logan, I.T., 1995. Ordnance Survey policy and practical implications for the revision of rural mapping. Technical meeting of the Photogrammetric Society, London, 2t March 1995. (To appear in Pbotogrammetric Record, 1996). Welch, R., 1992. Photogrammetry in transition - analytical to digital, Geodetical Info Magazine, 6(7): 39-41. Welch, R., 1993. Low-cost softcopy photogrammetry. Geodetical Info Magazine, 7(12): 55-57.
ISPRS Journal of Photogrammetry& RemoteSensing51 (1996) 188-195
Digital images in the map revision process P.R.T. N e w b y GEO-UK Ltd, 9 Merrytree Close, West Wellow, Romsey, Hants S051 6RB. UK
Accepted 17 January 1996
Abstract Progress towards the adoption of digital (or softcopy) photogrammetric techniques for database and map revision is reviewed. Particular attention is given to the Ordnance Survey of Great Britain, the author's former employer, where digital processes are under investigation but have not yet been introduced for routine production. Developments which may lead to increasing automation of database update processes appear promising, but because of the cost and practical problems associated with managing as well as updating large digital databases, caution is advised when considering the transition to softcopy photogrammetry for revision tasks.
1. Introduction It is clear from the current literature that photogrammetry is in the throes of a revolution (Welch, 1992; Heipke, 1995). The means now exist to leave behind the silver halide graphic images on which we have relied since the beginning of photogrammetry, and to manipulate instead a digital model of the world. Airborne sensors may eventually follow satellite-borne sensors into the digital data era. However, the state of the art of photogrammetry (as it affects the revision or updating of maps and of digital topographic databases) is still to make aerial photographs on polyester-based silver halide emulsions using cameras which would be broadly recognisable to the photographers of fifty years ago. The revolution so far only affects the subsequent stages of processing. As in all revolutions, forces of reaction also are at work, based in this case on sound economic and practical grounds. Thus, in this paper, no assumptions are made about the outcome. Instead, the state of development of map and database
revision towards the end of the 1992 to 1996 session of International Society for Photogrammetry and Remote Sensing (ISPRS) is reviewed, and some examples of promising research activities are cited,
2. The transition from analogue to digital photogrammetry Developments in computer science over the past thirty-five years have provided the driving force for new surveying, mapping, and photogrammetric techniques and equipment. The first productionised analytical aerial triangulation system which used electronic recording stereocomparators and "highspeed' computers largely mimicked an existing system based on the use of manual recording stereocomparators and hand calculators while greatly extending its possibilities (Arthur, 1959; Proctor, 1962). The subsequent astonishing developments in processing speeds and on-line storage, together with equally remarkable reductions in cost, have brought us through digital (or numerical) cartography, computer-assisted
0924-2716/96/$15.00 © 1996ElsevierScienceB.V. All rights reserved. Pll S0924-27 16(96)0001 2-3
PR.T. Newby /ISPRS Journal of Photogrammetry & Remote Sensing 51 (1996) 188-195
digital data capture using analogue stereo-plotters equipped with encoders, and later, analytical stereoplotters in which optical and mechanical parts were minimised and the photogrammetric restitution was carried out in real time by computational means. As numerical cartography gradually evolved into today's topographic databases and geographical information systems (GIS), graphical map output from photogrammetric plotters became less relevant. Moreover, those who already possessed topographic databases soon recognised that their concern in the future would be maintaining an up-to-date database. Thus, by the end of the 1980s the photogrammetric plotter had become a photogrammetric workstation at which a human operator, assisted by one or more computers, could capture or edit three-dimensional geometric, topological and semantic data about the world portrayed in collections of high-quality, twodimensional graphic images and recorded in digital databases (Bonjour and Newby, 1990). Meanwhile our colleagues developing the newly emerging discipline of remote sensing were employing digital images transmitted from satellites. Developments in computer science parallel to those exploited by surveyors and mappers allowed them to handle such images. Image processing and computer vision developed independently of photogrammetry for the simple reason that photogrammetrists retained the major advantage of working with graphic images of very high geometric and radiometric quality. Only very recently have developments in computer graphics made digital images worthy of the attention of photogrammetrists. Now there is a most welcome convergence between photogrammetry and remote sensing, image processing and computer vision. It is clear that photogrammetry can benefit from existing expertise in digital image processing, while contributing traditional strengths in the rigorous understanding and manipulation of the geometry of image formation. Digital (or softcopy) photogrammerry has now developed to the point where numerous vendors offer viable systems, and civilian mapping organisations in some of the world's more advanced countries have at least tried some aspects of digital photogrammetry, even if they have not yet adopted them whole-heartedly (Colomer and Colomina, 1994; Heipke, 1995). For map revision there are obvious possibilities
189
of using either the digital orthophoto as a data source for monocular data capture, or a digital photogrammetric workstation (DPWS) with stereoscopic viewing as a three-dimensional edit station. To this author's knowledge, however, there are as yet no organisations using softcopy photogrammetry as a routine production tool for map or database revision. Looking further ahead, researchers in several countries have made progress in automating change detection and hence feature extraction from digital images. Although there are some examples of this work, it is likely to be a long time before such techniques leave the research laboratory and enter the daily routine of those who earn their living by maintaining topographic databases. In the real world of routine production, all technical processes stand or fall by their cost effectiveness. This is especially the case with entirely new developments such as softcopy photogrammetry that involve high initial investments for research and development, capital equipment, staff training and (most importantly for database revision) integration with existing systems. To justify changes, cost benefits must be demonstrable. The following section will illustrate how this has worked in practice in a leading national mapping organisation. 3. Current map revision practice, with special reference to Great Britain
The large scale digital database in Great Britain is now complete. Approximately 60,000 sheets originally mapped at 1:1250 scale, some 160,000 at 1:2500 scale and almost 2000 at 1:10000 scale have been converted to about 220,000 digital files in which distinctions between old style map scales are becoming increasingly blurred. Moreover, all change on the ground of major importance to the map user is surveyed and the database is updated accordingly within six months of the change occurring. This represents a significant evolution of the practice of the Ordnance Survey of Great Britain (OS) in response to one of the world's most sophisticated communities of civilian customers for topographic information. Much has been written about OS revision practice in recent years (for example, Newby and Proctor, 1990; Newby, 1990a; Newby, 1994; Vincent and Logan, 1995). This section will provide a summary of
190
P.R.T. Newby/ISPRS Journal of Photogrammet,3, & Remote Sensing 51 (1996) 188-195
developments in both practice and policy, the result of technical research and development, consultation with customers, and the relentless pressure from government to reduce costs, increase efficiency, and involve private enterprise in its activities. For many years OS boasted of its system of continuous revision of major change, but in practice this nevertheless allowed substantial backlogs to build up, and minor change could take as much as forty years to be captured. The customers for OS digital data, or for the graphics derived from that data, are now more demanding, hence the emergence of the new system mentioned above, termed 'low threshold revision', by which major change reaches the database within six months of its occurrence on the ground. Reaching the present position involved a hectic catching-up exercise, most cost-effectively carried out using analytical photogrammetric workstations with superimposition of the existing digital data. In order to meet tight deadlines, revisions in many areas were done by field surveyors. What should in future constitute 'major change' has been the subject of lively debate as well as consultation with customers; however, most major change in future will surely be surveyed in the field because by definition 'low threshold' implies small increments. The exception will be the survey of major development projects of short duration, where timely photography will enable economic photogrammetric revision within the six month time frame. It is also possible that aerial photographs will be used as an aid to detection of urban change. Meanwhile minor change will move from being a continuous (but generally infrequent) process to a cyclic process at 5 year intervals (10 years in mountain and moorland areas), beginning with the clearance of outstanding backlogs. Thus, the new revision strategy which has emerged over the past four years can be summarised as follows: improve urban continuous revision, introduce cyclic revision for rural and mountain areas, and extend continuous revision of major change to rural areas. The cyclic revision of rural areas for minor change will be largely a photogrammetric process, both with regard to the detection and the survey of change. Change on the ground affects some 2% of detail in the database every year. Detecting that change, and verifying the absence of change from the
remaining 98%, both form part of the revision process, both incur significant costs, and neither can be neglected. With 160,000 map files for rural (1:2500 scale) areas, a five year cycle implies the revision of over 30,000 files per year. At any one time, hundreds or even thousands of files may be in work. Just managing these data is a demanding task. Data transfers, backups and validations require discipline, quite apart from the process of update itself, and the topological structure verification that must follow every update operation. This, of course, is for vector data. As soon as softcopy images are introduced, the volume of data will increase by up to three orders of magnitude as vector map files of typically well under a megabyte (Mb) are backed by images of the order of 100 Mb each. The processes used to survey change have been comprehensively documented. Newby (1990b) reported trials which led to the acceptance of the analytical 'plotter' with superimposition as the most cost effective means of updating OS large scale maps for a wide range of densities of change. However, all of the methods tested were found to be workable and were actually used over a protracted period for historic reasons associated with the legacy of earlier generations of equipment. These methods ranged from pure graphic plotting using analogue plotters, with subsequent digital data capture at a conventional digitising table, through on-line editing at a workstation linked directly to an upgraded analogue or an analytical plotter, to the current method of choice with the analytical plotter and superimposition. Here the operator's view through the eyepieces includes the three-dimensional landscape as portrayed on silver halide diapositives exposed using a modem camera with forward motion compensation, the old and new database (map) detail, and the menu for the editing process using LaserScan's LITES2 software. This is a cost-effective process which points towards a major advantage of softcopy systems where the superimposition of digital image and line-map detail from the existing database will come as part of the system at no extra cost. Although each successive development has smoothed the process of on-line updating, in practice a final edit is always done at a comparatively inexpensive two-dimensional graphics workstation away from the photogrammetric workstation itself.
I~R.Z Newbv /lSPRS Journal of Photogrammetry & Remote Sensing 51 (1996) 188-195
What now of the future? At the time of the last OS procurement of photogrammetric revision equipment, DPWS did already exist in the marketplace. It was considered premature to adopt them for database revision then, because their integration with existing data management practices would have delayed their effective contribution to clearing revision backlogs. Now, two years on, any organisation purchasing photogrammetric equipment for database revision should at least consider the numerous available DPWS. Meanwhile, OS has been conducting extensive developments and trials of digital monoplotting processes (Vincent and Logan, 1995). These use a digital (softcopy) aerial photograph obtained by scanning a conventional negative or diapositive, together with a digital terrain model (DTM) which enables the correct geometry to be extracted without the use of stereoscopic viewing. In principle either the raw digital image or a digital orthophoto may be used. The software for superimposing the existing data and updating the database with newlysurveyed change will obviously be simpler in the latter case. OS have disclosed that both Intergraph and LaserScan systems (the latter using DEC Alpha workstations) have been in use in the trials, but results have not been published. However, a recent public invitation to tender for a new photogrammettic scanner allows us to speculate that their digital photogrammetric trials, which date back over six years, will soon bear fruit in the production arena (Newby, 1990c). Since a great deal of OS revision work is done by field surveyors it is also worth mentioning the processes used. The amount of ground control depends on the extent of the area being revised and the nature of existing surrounding detail. If new control is required, it will be supplied by a suitable combination of total station traversing and Global Positioning System (GPS) surveys. Some new map detail may also be collected by total stations, but complete map data capture by electronic recording at the total station has not found as much favour among OS field revisers as among engineering project surveyors. Instead, a graphical survey method peculiar to OS is used to complete the new map detail, with or without the assistance of rectified enlargements of aerial photographs, followed by database update on a graphics workstation with digitising tablet in the field office.
191
This leads us to an exciting development in field surveying - at the opposite end of the scale from developments in GPS - the PRISM (Portable Revision and Integrated Survey Module) project based on a pen computer. This has the potential to become a complete, portable, interactive edit station for map data, replacing the graphic processes used hitherto. After several years of development PRISM is now in intensive production trials with some eighty units in use (Greenway, 1994; Vincent and Logan, 1995). If this proves to be a cost effective tool it will shift the economic balance between field and photogrammetric processes. Such a device could eventually incorporate data from total stations or GPS, or indeed carry a digital image backdrop as well as line map data, thereby placing a DPWS in the hands of the field surveyor and blurring historic distinctions between field and photogrammetric surveys! The government of Great Britain is increasing the role of the private sector in its map revision activities. Revision is now being done by a mixture of in-house work and private sector contracts, the latter primarily intended to assist rural cyclic revision. Intense competition in the market place and European Community rules on public contracts have led to contract awards to private sector companies not only from the United Kingdom (including Northern Ireland) but also from Denmark. It remains to be seen how the private sector, especially from outside Great Britain, copes with the peculiar disciplines of national database update. There is clearly a strong intent to make the process a success, and an enlightened application of quality systems and standards such as BS EN ISO 9000 and ISO 2859 in its support. In the absence of published information we can only speculate on the revision techniques. However, although product rather than method specifications are being used, it seems likely that contractors will be playing safe with well-trusted classical procedures rather than pushing back the frontiers with new developments.
4. Aerial images and their conversion to digital images It is not feasible to enter digital photogrammetry half-heartedly. To justify the investment in a scanner to convert photographic hard copy images into soft-
192
P.R.T. Newby / ISPRS Journal of PhotogrammetO' & Remote Sensing 51 (1996) 188-195
copy it is necessary to plan to make use of its full capacity. Unless reliance is to be placed on an agency service for photo scanning, a substantial investment in both a scanner and numerous DPWS at the same time is required. There has been extensive speculation on whether it is necessary to spend the large sums required for a top quality scanner designed expressly for photogrammetry, or whether much lower cost systems designed for less demanding tasks, or at least using off the shelf components, can give acceptable results. Certainly, it is now recognised that not all scanners are the same, that even very expensive scanners may not provide perfect results, and that the complete digital photogrammetric flowline will depend on this single crucial component (Bill, 1994). The importance of this issue has led to the formation of a joint European Organization for Experimental Photogrammetric Research (OEEPE)-ISPRS working group on the analysis of photo-scanners. Some key factors in the performance of scanners were propounded at an early meeting of that working group (K61bl et al., 1994). These covered: (1) the technical requirements of geometry, resolution, image noise, dynamic range and so on; (2) comfort and convenience aspects, such as the level of automation of the process and the use of original roll film negatives; and (3) engineering points involving the satisfactory solution of classical problems of aerial imagery such as optical distortion, image flatness, vibration and the avoidance of artefacts. Above all, the results of scanning must aim to be as good and cheap as a diapositive! This makes it quite hard for digital photogrammerry to compete. Photogrammetric operators have only just got used to the improvements which followed innovations in aerial photography such as new lens designs, forward motion compensation (FMC) and higher resolution emulsions - all of which have led photogrammetrists to expect sharper, better images than ever before. Moreover, for the surprisingly demanding task of detecting and surveying change while minimising the need for subsequent field completion, exploiting the improved geometric accuracy of modem photographs and measuring systems by reducing the scale of photography has not proved a sensible course. Instead, ease of interpretation has become the dominant requirement. Throwing all of this away in exchange for the privilege of viewing
lower resolution pixel-based images on a computer screen requires an act of faith or a dramatic cost advantage. Nevertheless, Stirling (1995) has reported that users of digital systems do adjust to the comparatively fuzzy screen images and that the loss of information is not severe. This introduces the question of the trade-off between pixel resolution and data volume which affects the cost of capture, storage and transmission of data. The standard was set by the Zeiss/Intergraph PhotoScan, which allows a smallest pixel size of 7.5 /~m. This implies about 600 Mb of data per aerial image. Much discussion of the minimum practical resolution has ensued, but no firm consensus has yet emerged. However it seems likely that most users will eventually settle for pixels in the 15 to 25 # m range, or some 100 Mb per image (Welch, 1993). Whether or not the next generation has difficulty with the storage volumes involved (terabytes per year for the national mapping of a medium sized country), processing speeds and transmission rates will certainly be critical, as will the disciplines of managing such quantities of data. Until very recently the idea of replacing the conventional aerial camera with a digital system in the aircraft seemed ludicrous. Developments over the past three years counsel caution in dismissing such ideas. The Institut Geographique National (IGN) (France) has been experimenting with airborne digital cameras and, through OEEPE, is seeking collaborators in order to develop the concept further. Meanwhile, in North America, systems are in use for environmental monitoring and are producing images which show some potential even for urban map revision (Monday et al., 1994). At the same time there are plans for commercial, satellite-borne, highresolution imaging systems as discussed by Konecny and Schiewe (1996) in this issue. With pixel sizes of one to two metres on the ground these could be of interest for small-scale map revision, but they will surely not replace airborne imagery for the update of large-scale databases. 5. DPWS hardware and software considerations At its most basic level the DPWS has monocular viewing only, of a single image whose geometry is corrected by a good DTM. Any monocular system
PR.T. Newby /ISPRS Journal of Photogrammet O, & Remote Sensing 51 (1996) 188-195
deliberately discards one of the most crucial benefits of photogrammetric surveying, namely stereoscopic viewing. Of course, this serious disadvantage may be outweighed by the obvious simplicity and economy of monoplotting systems. If not, it will be necessary to wait for the hand-held field computer to combine monocular photogrammetry with the surveyor's eye view at ground level before the loss of stereoscopy will be tolerated by most photogrammetrists. At the next level of present-day development, with the addition of stereoscopic viewing, the DPWS replicates the functionality of an analytical plotter. Like the old universal analogue machines, in its most advanced form it will be able to be used for all known photogrammetric processes. All DPWS will have the facility of viewing the existing vector or raster map data, together with any new c~ata captured from the current aerial image. For revision purposes the capture and edit system must integrate easily with the existing database or GIS. This integration points to a major advantage of digital photogrammetry in terms of quality management - real world geometry will always be maintained, or at least departures from it will be immediately apparent. Along the way the DPWS will be able to generate DTMs automatically and will allow them to be edited in a user-friendly way. From the DTM, orthophotos will in turn be generated automatically (Schiewe and Siebe, 1994). In order to use orthophotos of urban areas for map revision (or indeed any other purpose) it is essential that they show the tops, as well as the bases, of buildings in their correct planimetric positions. A solution to this obvious requirement, long overlooked, was described by Meister and Dan (1994) at the Commission IV Symposium held in Athens, Georgia. An up-dated report on the procedures is presented by Sinning-Meister et al. (1996) elsewhere in this issue. At least one leading vendor (Leica) now includes a similar procedure in a DPWS, although the onus remains on the user to capture the necessary three-dimensional building model manually in order to create a displacementfree orthophoto. Once the 'building lean' problem is eliminated automatically, the floodgates will open for the acceptance of orthophotos for urban mapping and database revision.
193
6. Other promised advances in map and database revision Almost all of the processes described so far have still relied on the involvement of the human operator. To contain the cost of database maintenance in a cost-conscious world, researchers are looking increasingly to automation. The excellence of the human eye and brain (or what soldiers call 'the Mark 1 eyeball') at tasks of image understanding and pattern recognition, however, is even harder to reproduce in a computer than it is to understand. Nevertheless, some progress has been made in recent years that points to the possibility of at least semi-automated processes reaching the production arena. One of the less demanding tasks in image understanding is to recognise and follow linear features such as roads. Successful demonstrations have been given using both SPOT and aerial image data (De Gunst et al., 1991: Solberg, 1992; Sakoda, 1993; Plietker, 1994; Peled, 1994). Comparison of a new image with the old network allows the latter to be updated. Peled has outlined a plausible scenario for future progress, beginning with a semi-automatic process of subtraction of old and new images followed by noise removal, to supply the human reviser with candidate areas for his attention. Next, in what he calls 'GIS-driven updating,' the above process would provide the trigger for automated recognition of new detail and its extraction in a hierarchy of themes. Finally, autonomous rule-based artificial intelligence (AI) systems may largely take over from the human operator. As in most aspects of image understanding, it seems unlikely that any one algorithmic approach will yield the required results, but that a combination of approaches, mimicking the human's intuitive combined tactics, may eventually be successful. Similar efforts have been reported in the detection and extraction of buildings, notably by McKeown et al. (1985) and Murakami and Welch (1992). Thus far, the distinction between two- and threedimensional databases has not been fully defined. A map models a three-dimensional world and, until recently, most GIS developers have preferred to treat it as two dimensional. Certainly the topological structure becomes very much more complex if the third dimension is allowed to intrude. Theoreticians have
194
P.R.T. Newby /ISPRS Journal of Photogrammetry & Remote Sensing 51 (1996) 188-195
now started to wrestle with the question of whether time ought to be treated as a fourth dimension in a GIS or whether update merely generates a succession of states of the database (which may or may not be stored for posterity). If the latter view is accepted, it makes it easier for us to consider practical matters such as change on the ground which has not yet reached the database, change in the database as a result of improved data without any corresponding change on the ground, and the requirements of users for complete new versions of the database after update or merely replacement of the updated features. As data structures become more rigorous and complex, these matters pose formidable problems of data integrity for both supplier and customer. They are necessarily addressed in practice by organisations like OS, but are also now receiving theoretical attention from researchers such as Kemppainen (1994).
7. Conclusions Digital or softcopy tools will undoubtedly be used for revision in the future but must compete effectively on cost grounds as well as being tailored to interact with existing database structures and practices, for this is the essence of revision as opposed to original survey. It is not just the storage but also the manipulation and transmission of the huge volumes of data entailed in digital images which will demand careful attention. The question of monoversus stereo-viewing is an important issue today, but in this author's opinion stereoscopy is a special advantage of photogrammetry which will not lightly be given up, at least until developments in data handling and transfer allow the field surveyor to hold a (monocular) DPWS in his hand, on site. Database revision has now become a respectable discipline within the ISPRS commission structure. Technical and institutional aspects rank equally as worthwhile subjects for discussion. There has been perceptible and promising progress towards the longterm goals of automated change detection and feature extraction but practical solutions are still a long way off. Database integrity after update is now attracting serious attention, being helped in practice by the use of the best tools and the best possible imagery as an aid to interpretation. The forthcoming Vienna Congress of ISPRS and the next session from 1996
to 2000 should provide interesting developments in this very important practical subject.
Acknowledgements The author was formerly Manager of Photogrammetry and Survey Computations at the Ordnance Survey of Great Britain. He acknowledges his debt to that organisation and to other former employers both for his photogrammetric experience and for the opportunity to begin serving ISPRS. However, he is now obliged to rely on published sources for reports on new developments. The opinions expressed in this paper, and any errors which may inadvertently have been included, are therefore entirely the responsibility of the author and have no connection with the views of the Ordnance Survey.
References Arthur, D.W.G., 1959. Recent developments in analytical aerial triangulation at the Ordnance Survey. Photogramm. Rec., 3(14): 112-124. Bill, R., 1994. Scanning of hardcopy images: achievements and limitations. In: Heipke and Mussio (Editors), ISPRS Commission, I Tutorial Notes, Acquisition, Characterization and Archiving of Digital Imagery, 62 pp. Bonjour, J.D. and Newby, ER.T., 1990. Architecture and components of photogrammetric workstations. Photogramm. Rec., 13(75): 389-405. Colomer, J.L. and Colomina, I., 1994. Digital photogrammetry at the Institut Cartogrhfic de Catalunya. Photogramm. Rec., 14(84): 943-956. Greenway, I., 1994. The use of pen computers for revision of large scale mapping. Presented to 20th FIG Congress, Melbourne, Commission 5, paper 505.4, 10 pp. De Gunst, M.E., Hart, C.S.L.A., Lemmens, M.J.EM. and Van Munster, R.J., 1991. Automatic extraction of roads from SPOT images. Proceedings of ISPRS and OEEPE joint workshop on updating digital data by photogrammetric methods, Oxford. OEEPE Official Publication 27: 131-140. Heipke, C., 1995. State-of-the-art of digital photogrammetric workstations for topographic applications. Photogramm. Eng. Remote Sensing, 61(1): 49-56. Kemppainen, H., 1994. Modelling update propagation in spatial databases. Int. Arch. Photogramm. Remote Sensing, 30(4): 625-633. K61bl, O. et al., 1994. Documents of the OEEPE-ISPRS Joint Workshop on the Analysis of Photo-Scanners, EPF-Lausanne. McKeown, D.M., Harvey, W.A. and McDermot, J., 1985. Rulebased interpretation of aerial imagery. IEEE Trans. Pattern Recog. Mach. Intell., PAMI-7: 570-585. Meister, M. and Dan, H., 1994. Processing of geographic data
P.R.T. Newby / ISPRS Journal of Photogrammetry & Remote Sensing 51 (1996) 188-195 for CAAD-supported analysis and design of urban areas. Int. Arch. Photogramm. Remote Sensing, 30(4): 433--440. Monday, H.M., Urban, J.S., Mulawa, D. and Benkelman, C.A., 1994. City of Irving utilizes high resolution multispectral imagery for NPDES compliance. Photogramm. Eng. Remote Sensing 60(4): 411-416. Murakami, H. and Welch, R., 1992. Automatic feature extraction for map revision. Int. Arch. Photogramm. Remote Sensing, 29(B4): 569-575. Newby, ER.T., 1990a. Digital map revision at the Ordnance Survey. Int. Arch. Photogramm. Remote Sensing, 28(4): 146154. Newby, ER.T., 1990b. Photogrammetric developments in the Ordnance Survey in 1990. Photogramm. Rec., 13(76): 561576. Newby, ER.T., 1990c. New initiatives in image digitising at the Ordnance Survey. Int. Arch. Photogramm. Remote Sensing, 28(3/2): 639-656. Newby, ER.T., 1994. Revision policy and practice for Great Britain's topographic database. Int. Arch. Photogramm. Remote Sensing, 30(4): 260-267. Newby, ER.T. and Proctor, D.W., 1990. Revision of large-scale maps at the Ordnance Survey. ISPRS J. Photogramm. Remote Sensing, 45: 137-151. Peled, A., 1994. Revision of digital maps and GIS databases. Int. Arch. Photogramm. Remote Sensing, 30(4): 268-272.
195
Plietker, B., 1994. Semi-automatic revision of street objects in ATKIS database DLM 25/1. Int. Arch. Photogramm. Remote Sensing, 30(4): 311-317. Proctor, D.W., 1962. Adjustment of aerial triangulation by electronic digital computers. Photogramm. Rec., 4( 19): 24-33. Sakoda, B., 1993. Computer-assisted tracing of faint roads on satellite images. Proc. ACSM/ASPRS Annual Convention, New Orleans, 2:311. Schiewe, J. and Siebe, E., 1994. Revision of cartographical databases using digital orthoimages. Int. Arch. Photogramm. Remote Sensing, 30(312): 724-731. Solberg, R., 1992. Semi-automatic revision of topographic maps from satellite imagery. Int. Arch. Photogramm. Remote Sensing, 29(134): 549-556. Stifling, D.M., 1995. Review of the use of digital workstations in the UK. Technical meeting of the Photogrammetric Society, London, 24 October 1995. (To appear in Photogrammetric Record, 1996.) Vincent, G.N. and Logan, I.T., 1995. Ordnance Survey policy and practical implications for the revision of rural mapping. Technical meeting of the Photogrammetric Society, London, 2t March 1995. (To appear in Pbotogrammetric Record, 1996). Welch, R., 1992. Photogrammetry in transition - analytical to digital, Geodetical Info Magazine, 6(7): 39-41. Welch, R., 1993. Low-cost softcopy photogrammetry. Geodetical Info Magazine, 7(12): 55-57.