Automated quality assessment of precast concrete elements with geometry irregularities using terrestrial laser scanning

AUTCON-02045; No of Pages 13 Automation in Construction xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Automation in Construction journal homepage: www.elsevier.com/locate/autcon

Automated quality assessment of precast concrete elements with geometry irregularities using terrestrial laser scanning Qian Wang a,b, Min-Koo Kim c, Jack C.P. Cheng a, Hoon Sohn b a b c

Department of Civil and Environmental Engineering, The Hong Kong University of Science and Technology, Hong Kong Department of Civil and Environmental Engineering, Korea Advanced Institute of Science and Technology, Daejeon, South Korea Department of Engineering, University of Cambridge, Cambridge, United Kingdom

a r t i c l e

i n f o

Article history: Received 23 October 2015 Received in revised form 21 March 2016 Accepted 25 March 2016 Available online xxxx Keywords: Terrestrial laser scanner Quality assessment Dimension estimation Precast concrete element Building information modeling Mirror-aided scanning

a b s t r a c t Precast concrete elements are popularly used and it is important to ensure that the dimensions of individual elements conforms to design codes. However, the current quality assessment of precast concrete elements is inaccurate and time-consuming. To address the problems, this study presents an automated quality assessment technique which estimates the dimensions of precast concrete elements with geometry irregularities using terrestrial laser scanners (TLS). While the scan data obtained from TLS represent the as-built condition of an element, a Building Information Modeling (BIM) model stores the as-design condition of the element. Taking the BIM model as a reference, the scan data are processed to estimate the as-built dimensions of the element. Experiments on a specimen demonstrated that the proposed technique can estimate the dimensions of elements effectively and accurately. Furthermore, a mirror-aided scanning approach, which aims to achieve reduced incident angles in real scanning environments, is proposed and validated by experiments. © 2016 Published by Elsevier B.V.

1. Introduction Precast concrete elements have been widely used in buildings and civil infrastructures. Compared to cast-in-place concrete, precast concrete brings higher quality control, reduced construction time, and environmental benefits [1,2]. Despite these advantages, structural failure of precast concrete systems can occur if the dimensional quality of individual components does not conform to design codes, as reported in [3]. Thus, it is of great significance to perform quality assessment on precast concrete elements before they are shipped to construction sites. Currently, the quality assessment of precast concrete elements primarily relies on manual inspection using traditional tools such as measuring tapes and following specified design codes. However, manual inspection suffers from two problems. Firstly, the result of manual inspection is inaccurate and unreliable [4]. Secondly, manual inspection is timeconsuming and costly, especially for large-size and complex structures. Hence, it is necessary to provide solutions which are able to perform quality assessment for precast concrete elements in an accurate and efficient manner. In recent years, terrestrial laser scanners (TLS) have gained popularity since they are able to acquire range measurement data at a high speed and high accuracy [5]. Due to these advantages, TLS has been used for the quality assessment of civil structures, including structure deformation measurement [6,7], and the identification and quantification of surface damages [8,9]. Some studies have applied TLS to estimate the dimensions of civil structures. Bosché [10] reported an

approach for calculating the as-built poses and dimensions of CAD model objects from the scan data of an industrial building's steel structure. The authors published previously a few works on quality assessment of precast concrete elements using laser scan data [11, 12]. Research reported in [11] developed a dimensional quality assessment technique for precast concrete elements with rectangularshape surfaces, and the developed technique was validated on labscale test specimens. Additional on-going research will enhance the developed technique and implement it on full-scale precast concrete panels. Research reported in [12] developed a framework for surface quality assessment of precast concrete elements by combining 3D laser scanning with building information modeling (BIM). However, the applicability of the existing research on dimension estimation as described in [11,12] is limited only to elements with simple geometries, such as straight columns or rectangular panels. No study has been conducted to estimate the dimensions of elements with geometry irregularities. In addition, photogrammetry has also been used in quality assessment, either stand-alone or combined with laser scanning. Scaioni et al. [13] proposed techniques to measure deformations of the transversal cross-section and the longitudinal profiles of tunnels using photogrammetry. Safa et al. [14] developed an automated measurement system to detect defects in piping fabrications using both laser scanning and photogrammetry. Riveiro et al. [15] also applied these two technologies to measure the minimum vertical underclearance and beam geometry during bridge inspections. Although photogrammetry is more economical and accessible compared

http://dx.doi.org/10.1016/j.autcon.2016.03.014 0926-5805/© 2016 Published by Elsevier B.V.

Please cite this article as: Q. Wang, et al., Automated quality assessment of precast concrete elements with geometry irregularities using terrestrial laser scanning, Automation in Construction (2016), http://dx.doi.org/10.1016/j.autcon.2016.03.014

2

Q. Wang et al. / Automation in Construction xxx (2016) xxx–xxx

to laser scanning, it suffers from a lower accuracy. Since the application discussed in this study requires high accuracy, laser scanning is adopted. This study proposes an automated quality assessment technique for estimating the dimensions of precast concrete elements with geometry irregularities, particularly focusing on the transverse sides of precast concrete bridge deck panels, using TLS. The uniqueness of this study includes, (1) the developed quality assessment technique can automatically estimate the dimensions of irregular geometry elements, (2) a new edge line estimation algorithm is developed which can estimate edge lines with arbitrary orientations, and (3) a mirror-aided scanning approach is proposed and validated to achieve reduced incident angles in real scanning environments. This paper is organized as follows. Section 2 introduces the research background on precast concrete bridge deck panels and TLS. Then, the proposed quality assessment technique is described in Section 3. Section 4 validates the effectiveness and accuracy of the proposed technique using scan data of a laboratory specimen. Furthermore, a mirror-aided scanning approach is proposed and validated in Section 5. Lastly, Section 6 concludes the whole study and suggests future work. 2. Research background

Table 1 Necessary dimensions for quality assessment of the transverse sides of panels and their tolerances. Dimensions

Notations

Tolerances

Panel depth

d a1 (outer horizontal) a2 (outer vertical) a3 (inner horizontal) a4 (inner vertical) b1 (horizontal) b2 (vertical) c1 (horizontal) c2 (vertical)

−3 mm, +6 mm

Dimensions of shear keys

Locations of shear keys Locations of flat ducts

±6 mm b1: ±6 mm b2: ±3 mm c1: ±6 mm c2: ±3 mm

Note that the dimensions of a shear key include outer horizontal, outer vertical, inner horizontal, and inner vertical dimensions, denoted as a1, a2, a3, and a4, respectively. Locations of a shear key include both horizontal and vertical locations, denoted as b1 and b2, respectively. The horizontal location is defined as the horizontal distance from the center of the shear key to the left edge of the panel, while the vertical location is defined as the vertical distance from the center of the shear key to the top edge of the panel. Additionally, the locations of a flat duct, denoted as c1 and c2, are defined similar to the locations of a shear key.

2.1. Precast concrete bridge deck panels 2.2. Terrestrial laser scanners A complete precast bridge deck consists of a series of bridge deck panels. To connect adjacent panels and ensure the integrated performance of complete decks, two structural features, shear keys and ducts, are usually provided on the transverse sides of the bridge deck panels. Fig. 1 shows a transverse side of a precast concrete bridge deck panel, with a polygonal outer boundary, as well as shear keys and ducts on the panel. Shear keys are distributed along the transverse side and serve as transverse panel-to-panel joints. They are designed to eliminate relative vertical movement between adjacent panels and to transfer the traffic load from one panel to the next [16]. Shear keys are mainly in two categories, non-grouted match-case shear keys using epoxy adhesive and grouted female-to-female shear keys, and the one shown in Fig. 1 is the latter. Similarly, ducts are also distributed along the transverse sides. They are used to place post-tensioned longitudinal reinforcements, which put the transverse panel-to-panel joints under compression [1]. Ducts are usually rounded or flat in shape, and the one shown in Fig. 1 is a flat duct. It was reported that one main problem of bridge deck systems is the deterioration associated with transverse joints [16]. Therefore, the dimensions of the transverse sides of a panel should conform to specified design codes in order to guarantee the performance of transverse joints. PCI [1,17] specifies the necessary dimensions for quality assessment of the transverse sides of panels and their tolerances, as shown in Table 1. Fig. 2 illustrates these dimensions using an example of a panel, on which two identical shear keys and a flat duct are provided.

TLS measures the distance to a target by emitting laser beams and detecting the reflected signals from the target. When TLS is in operation, the scanner head keeps rotating vertically and horizontally so that TLS can measure the distances of different measurement points. The scan data obtained from each measurement point contain a set of three-dimensional X, Y, and Z coordinates, a row index (corresponding to the scanner head's vertical rotation), and a column index (corresponding to the scanner head's horizontal rotation). TLS mainly adopts two different techniques for distance measurement, time-of-flight, and phase-shift. TLS using the time-of-flight technique emits a laser pulse and measures the travelling time of the reflected pulse. Since the velocity of the laser is known, the distance measurement can be inferred from the travelling time. TLS using the phase-shift technique emits an amplitude modulated continuous wave and measures the phase shift between the emitted and reflected signals. The distance measurement can be obtained based on the phase shift and the wavelength of the modulated continuous wave. Between these two techniques, the time-of-flight technique has a longer measurement distance while the phase-shift technique has higher measurement speed and higher accuracy [18]. Due to the substantial advantages over conventional range sensors, TLS has been widely used in different applications include as-built model reconstruction [19–22], heritage conservation [23,24], earthwork volume measurement [25,26], construction progress tracking [27,28], structural health monitoring [29,30], etc. While the scan data obtained from TLS represent the as-built conditions of objects, the corresponding as-design objects are usually stored in BIM models, which represent a digital representation of physical and functional characteristics of a facility. Some studies have compared the scan data with BIM models for quality inspection of as-built conditions [12,31,32]. 3. Development of TLS-based automated quality assessment technique

Fig. 1. A transverse side of a precast concrete bridge deck panel with shear keys and flat ducts.

The proposed TLS-based quality assessment technique focuses on the transverse sides of bridge deck panels, which have a polygonal outer boundary, while a few shear keys and flat ducts are distributed on it, as shown in Fig. 3. The scan data are acquired by a TLS, as shown in Fig. 4(a), and represent the as-built geometry of the panel. Note that two different approaches for scan data acquisition are used in this

Please cite this article as: Q. Wang, et al., Automated quality assessment of precast concrete elements with geometry irregularities using terrestrial laser scanning, Automation in Construction (2016), http://dx.doi.org/10.1016/j.autcon.2016.03.014

Q. Wang et al. / Automation in Construction xxx (2016) xxx–xxx

3

Fig. 2. Illustration of necessary dimensions for quality assessment of the transverse sides of panels.

study, a direct scanning approach used in Section 4 and a mirror-aided scanning approach discussed in Section 5. On the other hand, a BIM model is provided to represent the as-design geometry of the panel. Taking the BIM model as a reference, the scan data are processed through 5 steps, (1) data classification, (2) coordinate transformation, (3) extraction of inner corners, (4) extraction of outer corners, and (5) dimensions estimation, as illustrated in Fig. 4(b). Here, inner corners, shown as dots in Fig. 3, include corners along the inner and outer boundaries of shear keys and corners along the outer boundaries of flat ducts; whereas outer corners, shown as crossings in Fig. 3, include corners along the outer boundary of the panel. 3.1. Data classification Once the scan data are acquired, data classification is performed to extract data points belonging to the intended target surface. Fig. 5 illustrates the three categories of scan data, where P1 and P2 are valid points since they belong to the target surface, P4 and P5 are background points since they belong to the background surface, and P3 is a mixed pixel. A mixed pixel occurs when the laser beam is split into two parts and lies on both the target surface and the background surface, respectively. While both reflected signals from two surfaces are received by the TLS, the resulting scan data point can be anywhere along the line of the laser beam [33]. Wherever a mixed pixel is, it usually has a larger spacing to its adjacent point. For example, D2 , 3 is greater than D1 , 2, and D3,4 is greater than D4,5, given that Di, j represents the distance between Pi and Pj. Data classification aims to remove background points and mixed pixels while retain valid points. Background points are relatively easy to remove, and the main challenge is removal of mixed pixels. Tang et al. [34] compared a number of mixed pixels detection algorithms with a set of experiments. Experimental results show that, no algorithm performs exceptionally well and different algorithms may be applicable to different environments. However, these algorithms focus only on the detection of mixed pixels, not on removal of both background points and mixed pixels. Therefore, this study adopts a density-based clustering algorithm, namely, Density-Based Spatial Clustering of Applications with Noise (DBSCAN) [35], which not only can detect and distinguish mixed pixels, but also can remove the mixed pixels and background points simultaneously. DBSCAN operates based on two parameters (ε and minPts) and three fundamental concepts (core point, directly density-reachable, and density-reachable). The three fundamental concepts are built on the parameters ε and minPts. A point p is a core point if there are at least minPts

points surrounding p at a distance of ε or less from p. A point q is directly density-reachable from p if p is a core point and the distance between q and p is not greater than ε. Point q is density-reachable from p if there exists a sequence (p1 , p2 …pn) of points with p1 =p and pn = q, where each pi+1 is directly density-reachable from pi. As the DBSCAN algorithm visits each point in a dataset, if one point is a core point, a new cluster is started and all the points that are density-reachable from the core point, as well as the core point itself, are included in the cluster; however, if one point is neither a core point nor density-reachable from any core point, it is labeled as noise. Fig. 6 shows an example of scan data, where the filled dots represent valid points, empty solid dots represent background points, and empty dashed dots represent mixed pixels. Assuming that all the valid points and background points are evenly distributed and they have the same distance between two vertically or horizontally adjacent points, denoted as d0. DBSCAN is applied to the scan data with minPts = 8 and ε = d1, which is the distance between two adjacent valid points in the diagonal direction (e.g. point A and point B). As a result, point A becomes a core point and a cluster, which will eventually include all the valid points, is started. Since point D is a mixed pixel, the distance between point C and point D, denoted as d2, is greater than ε. Therefore, point D is not density-reachable from A and not included in this cluster. After applying DBSCAN, all the valid points are included in one cluster, all the background points are included in another cluster, and the mixed pixels are labeled as noise. In reality, the distance between two adjacent valid points d0 varies and depends on the distance from the TLS to the target surface and the incident angle of the laser beam. Therefore, to ensure that all valid points are included in the same cluster, ε should be equal to or greater than the theoretical maximum distance, dmax, between any two adjacent valid points in the diagonal direction. Moreover, since the target surface may not be perfectly smooth, the actual value of d0 can be greater than the theoretical value. Therefore, a safety factor is considered and ε is set as 1.2dmax in this study. Once all the clusters are found from the scan data, the cluster of data points that have the least average distance to the TLS are taken as valid points, because background points are always farther away from the TLS. These valid points belong to and represent the target surface, thereby named the as-built object. 3.2. Coordinate transformation Once the as-built object is extracted, coordinate transformation is conducted to ensure the as-built object best matches with the asdesign object. To this end, a 3D transformation of the as-built object

Fig. 3. Illustration of outer boundary, inner corners, and outer corners.

Please cite this article as: Q. Wang, et al., Automated quality assessment of precast concrete elements with geometry irregularities using terrestrial laser scanning, Automation in Construction (2016), http://dx.doi.org/10.1016/j.autcon.2016.03.014

4

Q. Wang et al. / Automation in Construction xxx (2016) xxx–xxx

Fig. 4. Overview of the proposed quality assessment technique: (a) Schematic of TLS scanning, (b) Quality assessment procedures.

needs to be computed. However, since the target object is a quasi-2D object, the problem of finding a 3D transformation can be simplified to finding a 2D transformation as follows. (1) Find the fitted planes of the as-built (Fig. 7(a)) and as-design (Fig. 7(b)) objects. The fitted plane of the as-built object is obtained using all the scan data based on the least squares method. Since the as-design object is expressed by a BIM model, sampling points are generated on all the surfaces of the BIM model with identical sampling density, and the least squares fitted plane of sampling points is used instead. (2) Transform the as-built object so that its fitted plane overlaps that of the as-design object. This plane is then defined as the X–Y plane of a new Cartesian coordinate system. As shown in Fig. 7(b), the origin of the coordinate system is positioned at one corner of the object, and the X and Y axes overlap with the two boundaries of the object, respectively. After the simplification, the 2D coordinate transformation (a combination of translation along the X axis, translation along the Y axis, and rotation about the Z axis) for best matching is computed as follows. (1) Project the as-built and as-design objects onto the X–Y plane. (2) Extract the boundaries of the as-built and as-design objects (Fig. 7(c)). The boundary of the as-built object is represented by a set of points, which are obtained by finding the first and last data points in each row and column of the as-built object since laser scan data are arrayed in rows and columns. The boundary of the as-design object is represented by a set of line segments, which compose its outer boundary. (3) Run the iterative search algorithm. In each iteration, a different 2D coordinate transformation is generated and applied to the as-built boundary. Afterwards, for each point in the as-built boundary, find the least distance to the

Fig. 5. Classification of three categories of scan data: valid points (P1 and 2), background points (P4 and P5), and mixed pixels (P3).

as-design boundary. After finding the least distances for all points in the as-built boundary, calculate the mean square of all the distances. When the mean square is minimized, the coordinate transformation for best matching is found, as shown in Fig. 7(d). Finally, the as-built object is transformed by this coordinate transformation to best match the as-design object. 3.3. Extraction of inner corners The locations of inner corners, including corners along the inner and outer boundaries of shear keys and corners along the outer boundaries of flat ducts, are extracted through two steps, (1) surface registration, and (2) extraction of surface intersection points, as described below. 3.3.1. Step 1: Surface registration All the scan data of the as-built object are registered to surfaces of the as-design object by surface registration. Surface registration mainly works for the areas surrounding the shear keys and flat ducts because each shear key or flat duct has multiple surfaces nearby. In the following explanations, a shear key is taken as the example. As shown in Fig. 8(a)(b), a typical shear key has 6 surfaces (labeled from 1 to 6) and 8 corners (labeled from A to H) surrounding it. Since the as-built and as-design objects are best matched in the coordinate transformation, an as-built scan point should be registered to an as-design surface close to it. First screening is performed by establishing a distance threshold for each surface. Only when the distance from a point to the surface is less than the threshold, i.e. a point is inside the margin of the surface, the point can be registered to the surface

Fig. 6. Data classification based on DBSCAN.

Please cite this article as: Q. Wang, et al., Automated quality assessment of precast concrete elements with geometry irregularities using terrestrial laser scanning, Automation in Construction (2016), http://dx.doi.org/10.1016/j.autcon.2016.03.014

Q. Wang et al. / Automation in Construction xxx (2016) xxx–xxx

5

Fig. 7. Coordinate transformation. (a) The as-built object represented by valid points. (b) The as-design object represented by a BIM model. (c) The boundaries of the as-built (dots) and asdesign (line segments) objects. (d) Coordinate transformation for best matching.

(Fig. 8(c)). For example, point P (Fig. 8(d)) inside the margin of surface 5 (area surrounded by dashed lines) is registered to surface 5. On the other hand, if a point is inside the margins of more than one surface, e.g. point Q (Fig. 8(d)) inside the margins of surfaces 1 and 5, additional criteria are needed to decide which surface this point should be registered to. Two measures are defined for registering as-built scan data, which are (1) distance measure and (2) differential angle measure. First, the distance measure is defined as the perpendicular distance from the point to each surface. As shown in Fig. 9(a), the distance measures of point Q to surfaces 1 and 5 are denoted as DQ1 and DQ5, respectively. Second, the differential angle measure is defined as the angle difference between a local plane around the point and each surface plane. As shown in Fig. 9(b), the local plane around point Q is extracted as the least

squares fitted plane of totally 9 points, including point Q and its 8 nearest neighbors. The normal vector of the local plane around point Q is denoted as nQ, while the normal vectors of surfaces 1 and 5 are denoted as n1 and n5, respectively. Thus, the differential angle measure between point Q and each surface is calculated as the angle difference between nQ and n1 or n5, and denoted as αQ1 and αQ5, respectively. Based on these two measures, the criterion for registering as-built scan data is defined as follows. When a surface has both the least distance measure and the least differential angle measure, the point is registered to this surface. Otherwise, the point is not registered to any surface. For point Q, if DQ1 b DQ5 and αQ1 b αQ5, point Q is registered to surface 1; if DQ1 N DQ5 and αQ1 N αQ5, point Q is registered to surface 5; otherwise, point Q is not registered to any surface. 3.3.2. Step 2: Extraction of surface intersection points Fig. 10 shows the registration of the as-built scan data to surfaces of a shear key, where points in the same color are registered to the same surface. Based on the registration result, all as-built surfaces (6 surfaces labelled from 1 to 6) are obtained as the least squares fitted planes of all the as-built points, which are registered to each surface. Then, the locations of all as-built corners (8 corners labelled from A to H) are extracted as the intersection points of three as-built surfaces. 3.4. Extraction of outer corners The locations of outer corners, which are along the outer boundaries of a panel, are extracted through two steps: (1) estimation of as-built edge lines, and (2) extraction of edge line intersection points. 3.4.1. Step 1: Estimation of as-built edge lines When estimating an edge line based on scan points close to the edge, there is an edge loss problem resulted from the removal of mixed pixels. Tang et al. [36] proposed a compensation model for edge loss specifically for horizontal or vertical edge lines. In this study, a new edge line estimation algorithm is proposed for estimating edge lines with arbitrary orientations.

Fig. 8. Registration of as-built scan data to surfaces of a shear key. (a) 3D view of a shear key. (b) Plane view of a shear key. (c) Margins of surfaces 1 (area surrounded by solid lines) and 5 (area surrounded by dashed lines). (d) Representative points P and Q.

(1) Extraction of edge points fully inside the target surface (type I edge points): Fig. 11(a) shows an example of scan data inside the target surface, which are represented as solid dots, including both empty and filled ones. Each point is labeled by a row index with numbers from 1 to 5, and a column index with letters from A to E. Thus, each point can be represented by its location, such as (2, B), which refers to the point at row 2 and column B.

Please cite this article as: Q. Wang, et al., Automated quality assessment of precast concrete elements with geometry irregularities using terrestrial laser scanning, Automation in Construction (2016), http://dx.doi.org/10.1016/j.autcon.2016.03.014

6

Q. Wang et al. / Automation in Construction xxx (2016) xxx–xxx

Fig. 9. Two measures for registering as-built scan data to surfaces. (a) Distance measure of point Q to each surface. (b) Differential angle measure between a local plane around point Q and each surface plane.

For the inclined true edge line, five points, including (1, D), (2, D), (3, C), (4, C), and (5, B) shown as solid filled dots, are extracted as type I edge points because they are the valid points closest to the true edge line in each row or column. Whether to select the closest points in each row or in each column as type I edge points depends on how many closest points can be found. In this example, if the closest points in each row are selected, the abovementioned five points will be found. However, if the closest points are selected in each column, only two points will be found, including points (4, C) and (2, D). Note that the last points in column A and B, i.e. points (5, A) and (5, B), cannot be regarded as type I edge points of the inclined edge because these two columns of points do not intersect with this edge line. Therefore, finding the valid scan points closest to the true edge line in each row rather than in each column is a better choice in this example. (2) Creation of edge points partially or fully outside the target surface (type II edge points): For each type I edge point, a corresponding type II edge point is created just next to it, by assuming that distance between two adjacent points in the same row or column remains the same. For example, point (1, E) is obtained by assuming that the distance between points (1, C) and (1, D) is the same as the distance between points (1, D) and (1, E). Thus, totally five type II edge points are created and they are denoted as dashed empty dots in Fig. 11. Note that, depending on whether these type II edge points are partially or fully outside the target

surface, they can be classified as either a mixed pixel or a background point. (3) Estimation of the edge line: As illustrated above, type I edge points are fully inside the target surface and type II edge points are partially or fully outside the target surface. Thus, the nearest endpoints of type I edge points to the edge line, represented as crossings in Fig. 11(a), are inside the target surface; the farthest endpoints of type II edge points to the edge line, represented as triangles in Fig. 11(a), are outside the target surface. Therefore, the true edge line should be able to separate these two groups of endpoints. Taking these two groups of endpoints as two classes of data, support vector machines (SVM) can be adopted to find the optimal separation line that has the maximum distance to any class of data [37]. However, in fact, the nearest or farthest endpoints of type I and type II edge points are unknown until the orientation of the edge is known. Therefore, the edge line is estimated as follows: (i) Take the centers of type I and type II edge points as two classes of data, and find the optimal separation line of these two classes of data using SVM, which is denoted as l1 in Fig. 11(b). (ii) Find the normal vector of l1, which is denoted ! ! as n . Note that the direction of n is always from the target sur! face to its outside. (iii) Translate l1 along the direction of n by a laser spot radius r so that l2 is obtained and taken as the estimated as-built edge line.

The accuracy of edge line estimation using the proposed algorithm is related to the spacing (d) of scan data and the scanning angle with respect to the edge line. Fig. 12(a) illustrates that the theoretical maximum error equals to d/2 for a vertical edge, when type I edge points are touching the true edge line. Therefore, as the spacing d increases, the estimation error increases proportionally. Fig. 12(b) shows that, as the angle θ between the scanning direction and the edge line increases pffiffi up to 45°, the theoretical maximum error decreases to 42d (¼ 2d cos45). 3.4.2. Step 2: Extraction of edge line intersection points Once all the as-built edges lines are extracted by the newly proposed edge line estimation algorithm, the location of each as-built corner is obtained as the intersection point of two edge lines. 3.5. Dimension estimation

Fig. 10. Registration of the as-built scan data to surfaces of a shear key.

As shown in Fig. 13, dots and crossings represent inner and outer corners of a panel which are extracted as described in Sections 3.3 and 3.4, respectively. Based on the locations of these corners, the necessary dimensions for quality assessment are estimated, including panel depth, dimensions of shear keys, locations of shear keys, and locations of flat ducts. The panel depth d is estimated at both ends of the panel, i.e. DO1O2 and DO3O4, given that DP1P2 represents the distance between P1 and P2. The dimensions of the shear keys are estimated as the average values

Please cite this article as: Q. Wang, et al., Automated quality assessment of precast concrete elements with geometry irregularities using terrestrial laser scanning, Automation in Construction (2016), http://dx.doi.org/10.1016/j.autcon.2016.03.014

Q. Wang et al. / Automation in Construction xxx (2016) xxx–xxx

7

Fig. 11. The proposed edge line estimation algorithm. (a) Extraction of two types of edge points from scan data—edge points fully inside the target surface (type I edge points) and edge points partially or fully outside the target surface (type II edge points). (b) Estimation of the as-built edge line.

Fig. 12. The theoretically maximum error in edge line estimation. (a) For a vertical edge. (b) For an inclined edge.

To examine the effectiveness and accuracy of the proposed quality assessment technique, a laboratory specimen was fabricated and experiments were conducted on the specimen. The scan data of the specimen were processed by the proposed technique, and the estimated dimensions were compared to the actual dimensions to evaluate the accuracy of dimension estimation.

100 × 100 mm and inner dimensions of 60 × 60 mm. Each flat duct had outer dimensions of 130 × 60 mm and inner dimensions of 90 × 20 mm. The geometry of the specimen is designed to represent the transverse view of the actual precast concrete panel. Although the specimen has a smaller size than the actual panel, they have similar geometric characteristics including a polygonal outer boundary, shear keys, and flat ducts. Fig. 15(a)–(b) show the experimental set-up and the actual concrete specimen. A FARO Focus 3D TLS was used to acquire the scan data of the specimen, which had an accuracy of ±2 mm at a distance of 20 m [5]. Multiple scans were conducted with varying scanning parameters, including (1) scanning distance between the TLS to the specimen, (2) angular resolution of the TLS, and (3) incident angle of the laser beam with respect to the specimen surface normal. A total of 27 scans were conducted, including scanning distances of 4, 6, and 8 m, incident angles of 0°, 15°, and 30°, and angular resolutions of 0.018°, 0.036°, and 0.072°.

4.1. Test specimen and experimental set-up

4.2. Data processing results

As shown in Fig. 14, the test specimen had polygonal outer boundary with dimensions of 1000 × 200 mm. Two shear keys and two flat ducts were provided on the specimen. Each shear key had outer dimensions of

The proposed quality assessment technique was applied to the scan data of the specimen. Firstly, data classification based on DBSCAN was implemented to extract data points representing the specimen. As

of a pair of distances. For example, the outer horizontal dimension of a shear key, denoted as a1, is estimated as (DI2I3 + DI1I4)/2. As for the locations of the shear keys, the center of a shear key is firstly obtained as the intersection point of the two diagonals of the outer corners. Then, the horizontal and vertical locations are estimated according to their definitions. Besides, the locations of a flat duct are estimated similar to the locations of a shear key. 4. Validation of the proposed technique

Fig. 13. Dimension estimation of a panel based on the previously extracted inner and outer corners.

Please cite this article as: Q. Wang, et al., Automated quality assessment of precast concrete elements with geometry irregularities using terrestrial laser scanning, Automation in Construction (2016), http://dx.doi.org/10.1016/j.autcon.2016.03.014

8

Q. Wang et al. / Automation in Construction xxx (2016) xxx–xxx

Fig. 14. Dimensions of the test specimen.

Fig. 15. Experimental set-up and test specimen. (a) Experimental set-up. (b) Actual concrete specimen.

shown in Fig. 16(a), the scan data were classified into several clusters with different colors. Among them, the cluster of points in blue color represents the test specimen. Then, the as-built object was transformed to best match the as-design object. As shown in Fig. 16(b), blue points represent the as-built object and the red line segments represent the outer boundary of the as-design object. Thirdly, the locations of inner corners were extracted through surface registration and extraction of the surface intersection points. Fig. 16(c) shows the results of surface registration and the locations of extracted inner corners shown in dots. Fourthly, the locations of outer corners were extracted through the proposed edge line estimation algorithm, shown as crossings in Fig. 16(d). Finally, the necessary dimensions for quality assessment were estimated, including panel depth, dimensions of shear keys, locations of shear keys, and locations of shear keys.

4.3. Accuracy of the estimated dimensions The estimated dimensions were compared to the actual dimensions, which were obtained by manually measuring the dimensions using a measuring tape with the smallest division of 1 mm. The errors between the estimated and the actual dimensions were computed and analysed as follows. Table 2 shows the errors of panel depth measurements with varying scanning parameters. Each scan had 2 panel depth measurements and Table 2 shows the average error and the standard deviation (in brackets) of these 2 measurements. Among all the 54 (27 × 2) original measurement errors, 51 of them (94.4%) were within the tolerance of panel depth (−3 mm, +6 mm). Note that when calculating the average error and its standard deviation, absolute values of errors were used.

Fig. 16. Quality assessment procedures of the test specimen using the proposed technique. (a) Data classification. (b) Coordinate transformation. (c) Extraction of inner corners. (d) Extraction of outer corners.

Please cite this article as: Q. Wang, et al., Automated quality assessment of precast concrete elements with geometry irregularities using terrestrial laser scanning, Automation in Construction (2016), http://dx.doi.org/10.1016/j.autcon.2016.03.014

Q. Wang et al. / Automation in Construction xxx (2016) xxx–xxx

9

Table 2 Average errors and standard deviations of panel depth measurements with varying scanning parameters. Average errors and standard deviations (in brackets) of panel depth measurements (mm) Incident angle (°)

Scanning distance = 4 m

Incident angle (°)

Angular resolution (°)

0 15 30

Scanning distance = 6 m

Incident angle (°)

Angular resolution (°)

0.018

0.036

0.072

1.5 (0.2) 1.6 (0.2) 1.6 (0.0)

1.7 (0.6) 2.4 (0.3) 2.8 (0.3)

3.2 (0.3) 3.6 (0.7) 4.1 (0.8)

0 15 30

Scanning distance = 8 m Angular resolution (°)

0.018

0.036

0.072

1.5 (0.1) 1.6 (0.2) 1.8 (0.3)

2.2 (0.4) 2.7 (0.1) 3.0 (0.7)

3.3 (0.6) 3.6 (1.0) 4.8 (0.7)

0 15 30

0.018

0.036

0.072

1.7 (0.2) 1.8 (0.4) 1.8 (0.4)

2.7 (0.6) 3.1 (0.6) 3.3 (0.8)

4.3 (0.9) 4.4 (1.0) 5.1 (1.2)

Table 3 Average errors and standard deviations of shear key dimension measurements with varying scanning parameters. Average errors and standard deviations (in brackets) of shear key dimension measurements (mm) Incident angle (°)

Scanning distance = 4 m

Incident angle (°)

Angular resolution (°)

0 15 30

Scanning distance = 6 m

Incident angle (°)

Angular resolution (°)

0.018

0.036

0.072

1.5 (0.3) 1.6 (0.5) 1.6 (0.7)

1.5 (0.5) 1.6 (0.5) 1.6 (0.6)

1.7 (0.5) 1.6 (0.7) 1.7 (0.8)

0 15 30

0.018

0.036

0.072

1.5 (0.4) 1.6 (0.5) 1.7 (0.5)

1.5 (0.7) 1.7 (0.6) 1.9 (0.7)

1.9 (0.6) 2.0 (0.7) 2.0 (0.6)

However, in the comparison with tolerances, the original values of errors with signs were used. Table 3 shows the errors of shear key dimension measurements with varying scanning parameters. Each scan had 8 measurements and Table 3 shows the average error and the standard deviation of the 8 measurements. Among all the 216 (27 × 8) original measurement errors, 216 of them (100%) were within the tolerance of shear key dimensions (±6 mm). Table 4 shows the errors of shear key location measurements with varying scanning parameters. Each scan had 4 measurements, including 2 horizontal location measurements and 2 vertical location measurements, and Table 4 shows the average error and the standard deviation of the 4 measurements. Among all the 54 (27 × 2) original measurement errors for horizontal locations, 54 of them (100%) were within the tolerance of shear key horizontal locations (± 6 mm). Among all the 54 (27 × 2) original measurement errors for vertical locations, 38

Scanning distance = 8 m Angular resolution (°)

0 15 30

0.018

0.036

0.072

1.6 (0.6) 1.7 (0.5) 1.9 (0.4)

1.8 (0.4) 1.9 (0.3) 2.2 (0.7)

2.3 (0.9) 2.3 (1.0) 2.5 (0.9)

of them (70.4%) were within the tolerance of shear key vertical locations (±3 mm). Table 5 shows the errors of flat duct location measurements with varying scanning parameters. Each scan had 4 measurements, including 2 horizontal location measurements and 2 vertical location measurements, and Table 5 shows the average error and the standard deviation of the 4 measurements. Among all the 54 (27 × 2) original measurement errors for horizontal locations, 54 of them (100%) were within the tolerance of flat duct horizontal locations (± 6 mm). Among all the 54 (27 × 2) original measurement errors for vertical locations, 36 of them (66.7%) were within the tolerance of flat duct vertical locations (±3 mm). Fig. 17(a)–(c) show the effects of different scanning distances, incident angles, and angular resolutions on the errors of estimated dimensions, respectively. In each graph, the errors of all the four estimated dimensions are presented separately. Note that for each scan-

Table 4 Average errors and standard deviations of shear key location measurements with varying scanning parameters. Average errors and standard deviations (in brackets) of shear key location measurements (mm) Incident angle (°)

Scanning distance = 4 m

Incident angle (°)

Angular resolution (°)

0 15 30

Scanning distance = 6 m

Incident angle (°)

Angular resolution (°)

0.018

0.036

0.072

1.4 (0.5) 1.4 (0.5) 1.5 (0.4)

1.7 (0.3) 1.8 (0.3) 2.4 (0.4)

2.6 (0.5) 2.7 (0.5) 3.2 (0.8)

0 15 30

Scanning distance = 8 m Angular resolution (°)

0.018

0.036

0.072

1.4 (0.4) 1.5 (0.5) 1.6 (0.6)

1.6 (0.5) 2.1 (0.7) 2.4 (0.5)

3.0 (0.6) 3.5 (0.9) 4.0 (1.1)

0 15 30

0.018

0.036

0.072

1.6 (0.6) 1.7 (0.4) 1.8 (0.5)

2.1 (0.6) 2.5 (0.8) 3.5 (1.0)

3.8 (1.1) 3.5 (1.0) 4.8 (1.2)

Table 5 Average errors and standard deviations of flat duct location measurements with varying scanning parameters. Average errors and standard deviations (in brackets) of flat duct location measurements (mm) Incident angle (°)

Scanning distance = 4 m

Incident angle (°)

Angular resolution (°)

0 15 30

Scanning distance = 6 m

Incident angle (°)

Angular resolution (°)

0.018

0.036

0.072

1.5 (0.4) 1.6 (0.6) 1.6 (0.3)

1.6 (0.7) 1.9 (0.6) 2.2 (0.4)

2.2 (0.8) 3.0 (0.6) 3.2 (1.0)

0 15 30

Scanning distance = 8 m Angular resolution (°)

0.018

0.036

0.072

1.6 (0.6) 1.7 (0.4) 1.8 (0.5)

1.6 (0.7) 2.6 (0.6) 3.0 (0.8)

2.9 (1.0) 3.1 (0.8) 3.4 (0.9)

0 15 30

0.018

0.036

0.072

1.6 (0.4) 1.9 (0.5) 1.9 (0.6)

2.0 (0.5) 2.7 (0.9) 3.3 (0.5)

3.6 (0.7) 4.0 (1.0) 4.6 (1.3)

Please cite this article as: Q. Wang, et al., Automated quality assessment of precast concrete elements with geometry irregularities using terrestrial laser scanning, Automation in Construction (2016), http://dx.doi.org/10.1016/j.autcon.2016.03.014

10

Q. Wang et al. / Automation in Construction xxx (2016) xxx–xxx

Fig. 17. Effects of different scanning parameters on the errors of estimated dimensions: (a) Effect of different scanning distances, (b) effect of different incident angles, and (c) effect of different angular resolutions.

ning parameter, e.g. scanning distance of 4 m, the errors of all 9 scans (3 incident angles × 3 angular resolutions) were averaged and presented. According to Fig. 17(a), errors of all the four estimated dimensions increased as the scanning distance increased. As the scanning distance increased from 4 to 8 m, the errors increased by 33% (shear key dimensions) to 42% (shear key locations). According to Fig. 17(b), errors

of all the four estimated dimensions increased as the incident angles increased. As the incident angle increased from 0° to 30°, the errors increased by 13% (shear keys dimensions) to 35% (flat ducts locations). According to Fig. 17(c), errors of all the four estimated dimensions increased as the angular resolution decreased. As the angular resolution decreased from 0.018° to 0.072°, the errors increased by 33% (shear key dimensions) to 153% (panel depth).

Fig. 18. Comparison of direct scanning and mirror-aided scanning. (a) Direct scanning. (b) Mirror-aided scanning.

Please cite this article as: Q. Wang, et al., Automated quality assessment of precast concrete elements with geometry irregularities using terrestrial laser scanning, Automation in Construction (2016), http://dx.doi.org/10.1016/j.autcon.2016.03.014

Q. Wang et al. / Automation in Construction xxx (2016) xxx–xxx

11

5. Mirror-aided scanning approach

Fig. 19. Working principle of the proposed mirror-aided scanning approach.

Among all the four estimated dimensions, the errors of the shear key dimensions were least affected by all the scanning parameters. One possible explanation is that the shear key dimensions were estimated based on the locations of the inner corners only, while all the other three dimensions were estimated based on the locations of the outer corners or both the outer and inner corners. Since the approaches to extracting the inner and outer corners are different, the effects of the scanning parameters are different. Different scanning parameters mainly affect the spacing d between two adjacent scan data (spacing = scanning distance × angular resolution / cos(incident angle)). In the extraction of the outer corners, edge lines estimation is the key to obtaining an accurate estimation of outer corners. As discussed in Section 3.4, the error of edge line estimation is proportional to d. On the other hand, in the extraction of the inner corners, surface registration is the key to obtaining an accurate estimation. When d increases, there will be less data points registered to each surface, thereby adversely affecting the accuracy of the inner corners. However, the accuracy of the inner corners will not be affected as much as that for the outer corners, as long as there are “enough” points on each surface. The problem is to know how many points are “enough” for accurate estimation of the inner corners, and further research is warranted for this topic. To conclude, the accuracy of the estimated dimensions by the proposed technique is affected by all three scanning parameters—scanning distance, incident angle, and angular resolution. To achieve more accurate estimated dimensions, scanning parameters should be set properly. Generally, a smaller scanning distance, smaller incident angle and higher angular resolution bring about higher accuracy. With a scanning distance of 4 m, incident angle of 0°, and angular resolution of 0.018°, the average errors of the four estimated dimensions were 1.5, 1.5, 1.4, and 1.5 mm, respectively, which were substantially smaller than the tolerances (±3 – ±6 mm). As a result, the experiments demonstrated that the proposed quality assessment technique can estimate the dimensions of precast concrete elements effectively and accurately.

As stated in Section 4, the accuracy of the estimated dimensions is affected by scanning parameters. However, it is sometimes difficult to set up the TLS so that scanning parameters are optimized in real scanning environments. Fig. 18 shows a common situation in precast concrete plants, where multiple panels are closely placed next to each other. In Fig. 18(a), where the TLS is set up in a traditional way to scan the target panel directly, the incident angle can be very large or a direct line-ofsight from the TLS to the transverse side of the panel may not even be secured. To address this situation, a mirror-aided scanning approach is proposed. As illustrated in Fig. 18(b), a mirror is placed close to the target panel. The laser beam emitted from the TLS is firstly reflected by the mirror, and then reaches the target panel. The incident angle of the laser beam can be minimized by adjusting the angle of the mirror. Fig. 19 illustrates the working principle of the proposed mirroraided scanning approach. Assuming that a laser beam emitted from the TLS is reflected by the mirror and hits point P of the target object, the travelling distances of the laser beam before and after mirror reflection become s1 and s2, respectively. Since the TLS uses phase shift measurement, it will measure a total distance of s1 + s2 in the direction of the emitted laser beam. Therefore, the mirror image of the true data point P will be placed at point P′, and points P′ and P are symmetrical with respect to the mirror. Similarly, all the scan data of the target object will be located at their mirror images, which are symmetrical with respect to the mirror. To recover the target object from its mirror image, a coordinate transformation is needed. Since the proposed quality assessment technique described in Section 3 already includes the coordinate transformation step, recovery of the target object can also be performed using the same algorithm. To examine the feasibility of the proposed mirror-aided scanning, experiments on a laboratory specimen were conducted. As shown in Fig. 20, a test specimen with dimensions of 700 × 120 mm was designed and it had four identical shear keys with outer dimensions of 80 × 80 mm and inner dimensions of 65 × 65 mm. Figs. 21–22 show the experimental set-up of the direct scanning and mirror-aided scanning experiments, respectively. For the direct scanning experiment, the TLS was located 4 m away from the specimen and the incident angle was 0°. For the mirror-aided scanning experiment, a mirror with dimensions of 1 × 0.3 m was fixed with two vertical anchors positioned on a table. The distance between the TLS and the mirror was 3 m and the distance between the mirror and the specimen was 1 m so that the total distance from the TLS to the specimen was identical to that of the direct scanning experiment. Besides, the mirror angle was adjusted so that the laser beam reached the test specimen with an incident angle of 0°. For both scanning experiments, the TLS used the same angular resolution of 0.018°. The scan data from both experiments were processed by the proposed quality assessment technique and the dimensions of the specimen were estimated. By comparing to the actual dimensions obtained from manual measurement, the errors of estimated dimensions were

700 mm

80 mm

120 mm

80 mm

15 mm

120 mm

15 mm

Fig. 20. Dimensions of the test specimen used for the mirror-aided scanning experiments.

Please cite this article as: Q. Wang, et al., Automated quality assessment of precast concrete elements with geometry irregularities using terrestrial laser scanning, Automation in Construction (2016), http://dx.doi.org/10.1016/j.autcon.2016.03.014

12

Q. Wang et al. / Automation in Construction xxx (2016) xxx–xxx

Fig. 21. Experimental set-up of the direct scanning experiment.

Fig. 22. Experimental set-up of the mirror-aided scanning experiment.

calculated and are shown in Table 6. For the direct scanning, the average error was 1.7 mm, whereas for the mirror-aided scanning, the average error was 2.0 mm, an increase of 18% but still smaller than the tolerances of these dimensions. It indicates that the proposed mirror-aided scanning approach can also yield accurate estimated dimensions, although the accuracy decreased to some degree. 6. Conclusions and future work This study presents an automated quality assessment technique to estimate the dimensions of precast concrete elements with geometry irregularities using TLS. The scan data of a precast concrete element are

Table 6 Average errors and standard deviations of the estimated dimensions for both the direct scanning and the mirror-aided scanning experiments. Average errors and standard deviations (in brackets) (mm)

Direct scanning Mirror-aided scanning

Panel depth

Shear key dimensions

Shear key locations

Average

1.4 (0.4) 1.8 (0.5)

2.0 (0.7) 2.5 (0.6)

1.6 (0.8) 1.8 (0.8)

1.7 2.0

obtained via a TLS and processed through five steps, including (1) data classification, (2) coordinate transformation, (3) extraction of inner corners, (4) extraction of outer corners, and (5) dimension estimation. To validate the proposed technique, a laboratory specimen was manufactured and scanning experiments were conducted on the test specimen with varying scanning parameters. The scan data of the specimen were processed by the proposed quality assessment technique to estimate the dimensions of the test specimen. By comparing the estimated dimensions with the actual dimensions, the accuracy of the estimated dimensions were calculated and analysed. It was shown that the accuracy of the estimated dimensions was affected by different scanning parameters. With the optimal scanning parameters, the estimated dimensions could have an average error of 1.5 mm, which was substantially smaller than the tolerances of the dimensions. The experiments on the test specimen demonstrated that the proposed technique can estimate the dimensions of precast concrete elements effectively and accurately. Furthermore, to help achieve better scanning parameters in real scanning environments, a mirror-aided scanning approach is proposed. The mirror-aided scanning approach enables the laser beam to be reflected by a mirror and then reach the target object with a reduced incident angle. Moreover, the mirror-aided scanning enables the laser beam to reach the target object even when there is no direct line-of-

Please cite this article as: Q. Wang, et al., Automated quality assessment of precast concrete elements with geometry irregularities using terrestrial laser scanning, Automation in Construction (2016), http://dx.doi.org/10.1016/j.autcon.2016.03.014

Q. Wang et al. / Automation in Construction xxx (2016) xxx–xxx

sight from the TLS to the target object. Experiments on a laboratory specimen demonstrated that the proposed mirror-aided scanning could also yield accurate estimated dimensions of precast concrete elements. There are some limitations in this study, which are potential topics for future work. (1) The proposed quality assessment technique has been validated only on a laboratory specimen. Additional experiments on large-size precast concrete elements are needed to further examine the effectiveness of the proposed technique. (2) The proposed mirroraided scanning is currently only applicable to small-size specimen due to the limited size of the mirror. Further investigations are needed to extend its applicability to large-size precast concrete elements.

[16]

[17] [18] [19]

[20]

Acknowledgment

[21]

This research was supported by a grant (13SCIPA01) from Smart Civil Infrastructure Research Program funded by Ministry of Land, Infrastructure and Transport (MOLIT) of Korea Government and Korea Agency for Infrastructure Technology Advancement (KAIA).

[22]

References [1] M. Alhassan, PCI State-of-the-Art Report on Full-Depth Precast Concrete Bridge Deck Panels, Precast/Prestressed Concrete Institute, 2011 (ISBN: 978-0-9846705-1). [2] J. Glass, The Future for Precast Concrete in Low-Rise Housing, British Precast Concrete Federation, Leicester, 2000 (ISBN: 0-9536773-0-3). [3] P.L. Popovic, R.C. Arnold, Preventing failures of precast concrete facade panels and their connections, Forensic Engineering, Proceedings of the Congress 2000, pp. 532–539. [4] B.M. Phares, G.A. Washer, D.D. Rolander, B.A. Graybeal, M. Moore, Routine highway bridge inspection condition documentation accuracy and reliability, J. Bridg. Eng. 9 (4) (2004) 403–413, http://dx.doi.org/10.1061/(ASCE)1084-0702(2004)9:4(403). [5] FARO Focus3D - FARO, Inc., - http://www.faro.com/en-us/products/3d-surveying/ faro-focus3d/overview. [6] S.J. Gordon, D.D. Lichti, Modeling terrestrial laser scanner data for precise structural deformation measurement, J. Surv. Eng. 133 (2) (2007) 72–80, http://dx.doi.org/10. 1061/(ASCE)0733-9453(2007)133:2(72). [7] O. Monserrat, M. Crosetto, Deformation measurement using terrestrial laser scanning data and least squares 3D surface matching, ISPRS J. Photogramm. Remote Sens. 63 (1) (2008) 142–154, http://dx.doi.org/10.1016/j.isprsjprs.2007.07.008. [8] M.J. Olsen, F. Kuester, B.J. Chang, T.C. Hutchinson, Terrestrial laser scanning-based structural damage assessment, J. Comput. Civ. Eng. 24 (3) (2010) 264–272, http:// dx.doi.org/10.1061/(ASCE)CP.1943-5487.0000028. [9] M.K. Kim, H. Sohn, C.C. Chang, Localization and quantification of concrete spalling defects using terrestrial laser scanning, J. Comput. Civ. Eng. 29 (6) (2014), http:// dx.doi.org/10.1061/(ASCE)CP.1943-5487.0000415. [10] F. Bosché, Automated recognition of 3D CAD model objects in laser scans and calculation of as-built dimensions for dimensional compliance control in construction, Adv. Eng. Inform. 24 (1) (2010) 107–118, http://dx.doi.org/10.1016/j.aei.2009.08. 006. [11] M.K. Kim, H. Sohn, C.C. Chang, Automated dimensional quality assessment of precast concrete panels using terrestrial laser scanning, Autom. Constr. 45 (2014) 163–177, http://dx.doi.org/10.1016/j.autcon.2014.05.015. [12] M.K. Kim, J.C. Cheng, H. Sohn, C.C. Chang, A framework for dimensional and surface quality assessment of precast concrete elements using BIM and 3D laser scanning, Autom. Constr. 49 (2015) 225–238, http://dx.doi.org/10.1016/j.autcon.2014.07.010. [13] M. Scaioni, L. Barazzetti, A. Giussani, M. Previtali, F. Roncoroni, M.I. Alba, Photogrammetric techniques for monitoring tunnel deformation, Earth Sci. Inf. 7 (2) (2014) 83–95, http://dx.doi.org/10.1007/s12145-014-0152-8. [14] M. Safa, A. Shahi, M. Nahangi, C. Haas, H. Noori, Automating measurement process to improve quality management for piping fabrication, Structures 3 (2015) 71–80, http://dx.doi.org/10.1016/j.istruc.2015.03.003. [15] B. Riveiro, H. González-Jorge, M. Varela, D.V. Jauregui, Validation of terrestrial laser scanning and photogrammetry techniques for the measurement of vertical underclearance and beam geometry in structural inspection of bridges, Meas. J.

[23]

[24]

[25]

[26]

[27]

[28]

[29]

[30]

[31]

[32]

[33]

[34]

[35]

[36]

[37]

13

Int. Meas. Confed. 46 (1) (2013) 784–794, http://dx.doi.org/10.1016/j. measurement.2012.09.018. J.M. Wacker, M.O. Eberhard, J.F. Stanton, State-of-the-Art Report on Precast Concrete Systems for Rapid Construction of Bridges (No. WA-RD 594.1), Washington State Department of Transportation, 2005. PCI Committee on Tolerances, 135-Tolerance Manual for Precast and Prestressed Concrete Construction, Precast/Prestressed Concrete Institute, 2000. G. Petrie, C.K. Toth, Terrestrial laser scanners, Topographic Laser Ranging and Scanning Principles and Processing 2009, pp. 87–128. P. Tang, D. Huber, B. Akinci, R. Lipman, A. Lytle, Automatic reconstruction of as-built building information models from laser-scanned point clouds: a review of related techniques, Autom. Constr. 19 (7) (2010) 829–843, http://dx.doi.org/10.1016/j. autcon.2010.06.007. I. Brilakis, M. Lourakis, R. Sacks, S. Savarese, S. Christodoulou, J. Teizer, A. Makhmalbaf, Toward automated generation of parametric BIMs based on hybrid video and laser scanning data, Adv. Eng. Inform. 24 (4) (2010) 456–465, http:// dx.doi.org/10.1016/j.aei.2010.06.006. F. Bosche, C.T. Haas, Automated retrieval of 3D CAD model objects in construction range images, Autom. Constr. 17 (4) (2008) 499–512, http://dx.doi.org/10.1016/j. autcon.2007.09.001. X. Xiong, A. Adan, B. Akinci, D. Huber, Automatic creation of semantically rich 3D building models from laser scanner data, Autom. Constr. 31 (2013) 325–337, http://dx.doi.org/10.1016/j.autcon.2012.10.006. W. Boehler, G. Heinz, A. Marbs, The potential of non-contact close range laser scanners for cultural heritage recording, Int. Arch. Photogramm. Remote. Sens. Spat. Inf. Sci. 34 (5/C7) (2002) 430–436. N. Yastikli, Documentation of cultural heritage using digital photogrammetry and laser scanning, J. Cult. Herit. 8 (4) (2007) 423–427, http://dx.doi.org/10.1016/j. culher.2007.06.003. J.C. Du, H.C. Teng, 3D laser scanning and GPS technology for landslide earthwork volume estimation, Autom. Constr. 16 (5) (2007) 657–663, http://dx.doi.org/10. 1016/j.autcon.2006.11.002. E.J. Jaselskis, Z. Gao, R.C. Walters, Improving transportation projects using laser scanning, J. Constr. Eng. Manag. 131 (3) (2005) 377–384, http://dx.doi.org/10.1061/ (ASCE)0733-9364(2005)131:3(377). S. El-Omari, O. Moselhi, Integrating 3D laser scanning and photogrammetry for progress measurement of construction work, Autom. Constr. 18 (1) (2008) 1–9, http://dx.doi.org/10.1016/j.autcon.2008.05.006. Y. Turkan, F. Bosche, C.T. Haas, R. Haas, Automated progress tracking using 4D schedule and 3D sensing technologies, Autom. Constr. 22 (2012) 414–421, http:// dx.doi.org/10.1016/j.autcon.2011.10.003. H.S. Park, H.M. Lee, H. Adeli, I. Lee, A new approach for health monitoring of structures: terrestrial laser scanning, Comput. Aided Civ. Infrastructure Eng. 22 (1) (2007) 19–30, http://dx.doi.org/10.1111/j.1467-8667.2006.00466.x. D. González-Aguilera, J. Gómez-Lahoz, J. Sánchez, A new approach for structural monitoring of large dams with a three-dimensional laser scanner, Sensors 8 (9) (2008) 5866–5883, http://dx.doi.org/10.3390/s8095866. P. Tang, E.B. Anil, B. Akinci, D. Huber, Efficient and effective quality assessment of asis building information models and 3d laser-scanned data, Paper presented at the Congress on Computing in Civil Engineering, Proceedings 2011, pp. 486–493, http://dx.doi.org/10.1061/41182(416)60. F. Bosché, E. Guenet, Automating surface flatness control using terrestrial laser scanning and building information models, Autom. Constr. 44 (2014) 212–226, http:// dx.doi.org/10.1016/j.autcon.2014.03.028. M. Hebert, E. Krotkov, 3D measurements from imaging laser radars: how good are they? Image Vis. Comput. 10 (3) (1992) 170–178, http://dx.doi.org/10.1016/02628856(92)90068-E. P. Tang, D. Huber, B. Akinci, A comparative analysis of depth-discontinuity and mixed-pixel detection algorithms, Paper Presented at the 3DIM 2007—Proceedings 6th International Conference on 3-D Digital Imaging and Modeling 2007, pp. 29–36, http://dx.doi.org/10.1109/3DIM.2007.5. M. Ester, H.P. Kriegel, J. Sander, X. Xu, A density-based algorithm for discovering clusters in large spatial databases with noise, Kdd, vol. 96, No. 34 1996, August, pp. 226–231. P. Tang, B. Akinci, D. Huber, Quantification of edge loss of laser scanned data at spatial discontinuities, Autom. Constr. 18 (8) (2009) 1070–1083, http://dx.doi.org/10. 1016/j.autcon.2009.07.001. C. Cortes, V. Vapnik, Support-vector networks, Mach. Learn. 20 (3) (1995) 273–297, http://dx.doi.org/10.1023/A:1022627411411.

Please cite this article as: Q. Wang, et al., Automated quality assessment of precast concrete elements with geometry irregularities using terrestrial laser scanning, Automation in Construction (2016), http://dx.doi.org/10.1016/j.autcon.2016.03.014