reconstruction of 3D building model towards of 3D city model development by using Close-Range Photogrammetric (CRP) technique and to verify the limita...

0 downloads 269 Views 514KB Size
Geoinformation Science Journal, Vol. 10, No. 1, 2010, pp: 49-60

INTEGRATION OF AERIAL AND CLOSE-RANGE PHOTOGRAMMETRIC METHODS FOR 3D CITY MODELING GENERATION Nor’Ainah Amat, Halim Setan & Zulkepli Majid UTM Photogrammetry & Laser Scanning Research Group, Faculty of Geoinformation Science and Engineering, Universiti Teknologi Malaysia (UTM), Malaysia Email: [email protected], [email protected], [email protected]

Abstract Aerial photogrammetric method is currently used to generate the 3D city model at level of detail two (LOD2). However, the method faced with the problem of mapping the small buildings due to scaling factor. Therefore, this research aim to produce 3D building model of small building by using Close-Range Photogrammetric (CRP) method. The generated 3D model were than integrate with the 3D city model generated from aerial images for simple visualization and applications. Keywords : 3D City Model, 3D Building, Photogrammetry, PhotoModeler, Visualisation 1.0





In 3D city model development, it is important to choose appropriate data and suitable method (Kobayashi, 2007). Aerial photogrammetry data mainly used in 3D city model development compared to the airborne laser scanning data. Aerial photogrammetry performs semi-automatic process in developing 3D city model; manual digitizing on aerial images and performs automatic process in height determination. The 3D city models consist of fundamental component such as Digital Terrain Model (DTM), building models, street space models, and green space models (Jurgen et al., 2006). Recently, the reconstruction of 3D building model in 3D city model development has attracted public interest to improved 3D building model in terms geometry and textures. In the earlier production of 3D city model by using aerial photogrammetry, Vermeij and Zlatanova (2001) had reconstructed the building model as a block model. The block model is extracted from digitizing the building shape from the roof view. Textures and roof is not visualized in the block model. In 2003, Flamanc, has improved the 3D building model by adding roof on top of the block model without textures. Then it follows by Kobayashi (2007), which make an improvement on the building model textures and geometry. However, there still has constraint to develop building model from aerial images such as the recognition of the small building in aerial images. Aerial photogrammetry is used for smallscale mapping and the images are taken from high position by airplane. The aerial images gives an advantage to covers wide area on the ground but lack in recognition of small ISSN 1511-9491© 2009 FKSG


Geoinformation Science Journal, Vol. 10, No. 1, 2010, pp: 49-60

building due to the resolution of aerial images. Figure 1 shows the small building in aerial images.

Small building

Figure 1 The difficulties to recognize small building in aerial image In addition, the facade and geometry of the building from aerial images is visible from the air. The architectural instalments such as windows, door or balconies are not visible in aerial images. Therefore, due to the circumstances of aerial images in the reconstruction of 3D building model, this research objectives are to improve the current photogrammetric technique in the reconstruction of 3D building model towards of 3D city model development by using CloseRange Photogrammetric (CRP) technique and to verify the limitation of Close-Range photogrammetric technique in the reconstruction of 3D building model due to the geometry and textures of the building facade.



Close-range photogrammetry methods are fundamental for aerial photogrammetry (Hallert, 1960). In general, close-range photogrammetry is a technique of representing and measuring 3D objects using data stored on 2D photographs, which are the base for rectification (Vesna, 2008). In order to obtain 3D information, the two photographs of the same objects are necessary. Close-range photogrammetry is a part of terrestrial photogrammetry but has dissimilarity in camera-to-object distance. In the close-range photogrammetry, the limitation of camera-to-object distance is less than 100m (Cooper and Robson, 2000). Close-range photogrammetry is mostly used for deformation measurement of structures, architectural mapping, modelling buildings, documentation of artefacts, reverse engineering purposes, or remodelling traffics accidents and crime investigation. Architectural and archaeological photogrammetry is the example of close-range photogrammetry application that is widely been used since 1960s (Dallas, 1996). In architectural and archaeological, the 50

Geoinformation Science Journal, Vol. 10, No. 1, 2010, pp: 49-60

close-range photogrammetry is used as data capture to provide 3D CAD drawing for structures of the buildings or monument. The accuracy of close-range photogrammetry in producing 3D model for architectural and archaeological gives the promising result and it is proven by Hanke and Ebrahem (1997). Due to the capabilities of close-range photogrammetry in 3D measurement of architectural, an experiment has been conducted to create 3D building model for 3D city model application. Low cost close-range photogrammetry technique is the aims to create 3D building model.



In this research, aerial photogrammetry data is still used in order to produce DTM or base of the 3D city model while the close-range photogrammetry is to perform the 3D building reconstruction. In addition, the others objects or entities that can be placed in 3D city model such as streets, trees, and parking lot can be extract from aerial photogrammetry. Figure 2 shows the flow of aerial photogrammetry task and close range photogrammetry task. Closerange photogrammetry is the main task that highlighted in this research where it is used to improve the method of 3D building reconstruction by aerial photogrammetry. Improvement AERIAL PHOTOGRAMMETRY



VECTORIZATION Digitizing building outline,3D point at ground.











Figure 2 Flow of Aerial Photogrammetry and Close Range Photogrammetry Task


Aerial Photogrammetry Processing

The aerial images processing are including stereo model processing, vectorization, layer classification and development of base model. Stereo model processing is the process to 51

Geoinformation Science Journal, Vol. 10, No. 1, 2010, pp: 49-60

orient two aerial images and it is performs by using DVP software and the stereo model produces stereo images. The stereo images are used in vectorization stages which it involves the digitizing the building outline and extracted the 3D points in the stereo images. The 3D points consists of local coordinate information (x,y) and height information (z). Then, layer classification is performing to separate the 3D point and building outline layer for base model development. In order to complete the base model for 3D city model, the DTM are generated from 3D points and the buildings outline are drape onto the DTM. Figure 3 shows the process of DTM generation and Figure 4 shows the building outline on DTM. Both of these processes are perform by using ArcScene Software

Figure 3 Digital Terrain Model (DTM) generation from 3D point height.

Figure 4 Building outline drape onto the DTM Since the buildings are important component in the creation of 3D city model, the temporary buildings model can be placed to the base model. The temporary buildings model are modelled by extruded the buildings outline. The geometry and the height of the buildings are based on the buildings outlines. At this stage, landscape effect can be added to the model to make the model more realistic. The material that used as the visual effect is orthophoto image. Orthphoto image are produced from the aerial images and it is overlay onto the DTM. Figure 5 shows the extruded buildings with the landscape effect from orthophoto images.


Geoinformation Science Journal, Vol. 10, No. 1, 2010, pp: 49-60

Orthophoto overlay

Extuded buildings

Figure 5 Extruded buildings with landscape effect from orthophoto images


Close-range Photogrammetry Data Capture and Processing

The SONY CYBERSHOT DSCF828 digital camera is used for CRP data collecting. Before capturing the image, the camera needs to be calibrated to define camera parameters in order to provide accurate measurement. The camera parameters are focal length, principal point, imaging scale, and lens distortion. In this study, the camera is calibrated by using calibration grid from PhotoModeler software. PhotoModeler is close-range photogrammetry software. The calibration grid is a pattern of dots and design for the Camera Calibrator in PhotoModeler. The calibration is started by captured eight calibration images with four camera position and one landscape and one portrait shot at each position. While taking the images of the buildings, the camera position between two cameras must be in good angular separation which is close to 90º. The camera position is important in generate the correct position for the 3D point. The poor camera position occurs when the position between the two cameras is close to each other. In addition, the images must be overlap as much as possible in order to reference the points in the images. This is because, PhotoModeler needs points marked in two or more images and the images taken side by side should contain the same object features and points. The images of the building are captured as many as well to covers the entire building facades. This work has captures 19 images to covers the building facades. When completed capture the images, the building image are processed using PhotoModeler 6.0.

3.2.1 PhotoModeler Processing The building images that acquired are processed using PhotoModeler software to create the 3D building model. Figure 6 shows the processing step to generate 3D building by using PhotoModeler 6.0.


Geoinformation Science Journal, Vol. 10, No. 1, 2010, pp: 49-60

Select mode of project : Point-based project

Import Image

Camera Assignment

Image Registration


3D building and adding scale Figure 6 The Processing Step to Generate 3D Building by Using Photomodeler PhotoModeler 6.0 consists of two modes of project; point-based project and shaped-based project. Point-based project is a process to marking the point on the photograph to create the 3D objects. The shaped-based model is a process to create the 3D objects by marking the edges of the shape into the photograph. In this study, the point-based mode is selected to process the data because this mode is basically used to create 3D for complicated object. The next step is to import the building images that had been captured into the PhotoModeler 6.0. When starting the project, at least two or three photographs can be imported to the software because it helps to reduce the chance of failure in processing. The camera description has to be assigned into the software by importing the camera calibration file that had been made before capturing the image. PhotoModeler needs the camera description to create a geometrical relationship between points on the photograph and points in 3D space. The image registration process started after importing the calibration file. Image registration is the process to mark the point or edges of the objects on two or more different photograph and it is called as image referencing. The referencing must be mark on same position between the two photographs. Figure 7 shows the referencing between the two images. At least six points has to be referenced on the photograph in order to register the images.


Figure 7 The referencing between the two images 54

Geoinformation Science Journal, Vol. 10, No. 1, 2010, pp: 49-60

The digitising process is performed in order to create the 3D building model. The line mode is used to perform the digitising process. Figure 8 shows the digitised images. The image registration and digitising step are repeated when importing the new image to the software.

Figure 8 The digitized images Once the 3D building completed digitised, the model are scaled and oriented into coordinate system by using Three Point Rotation menu in PhotoModeler. Three known coordinates on the building are assigned into the model. In order to build up the photo-textures on 3D building model, the facades of the building was identified using surface path tool. This tool allows extract the textures from images into the model. Afterwards, the 3D building model are imported to 3D DXF (dxf) AutoCAD file as the building geometry and 3D Studio (3ds) file as the building textures to integrate with base model.


Integration Aerial and Close-Range Photogrammetry

The integration between aerial and close-range photogrammetry are execute in ArcScene. The 3D DXF file created in PhotoModeler are import into the base model as a building polygon while the 3ds cannot import directly into the base model. The 3ds file is needed to convert to Multipatch (mdb) file as a requirement in ArcScene to display 3D textures. Multipatch is a 3D geometry used to represent the outer surfaces of features that occupy a discrete area or volume in 3D surface (ESRI,2008). Google Sketch-up version 6 is used to help convert the 3ds file into mdb file so that the textures can import to the ArcScene. Figure 9 shows the 3D building integrates with base model in ArcScene.

Figure 9 The 3D building integrates with Base Model in ArcScene 55

Geoinformation Science Journal, Vol. 10, No. 1, 2010, pp: 49-60




The 3D Building Model

The images of simple building had been captured to produce the 3D building model. In order to create the 3D building model, 19 images are taken but only 6 images are used to process and create the 3D model. Figure 10 show the results of the 3D building model in wireframe and Figure 11 shows the 3D building model with texture.

Figure 10 Wireframe 3D building model

Figure 11 3D Building Model with Texture


Visual Analysis

In this study, the LoD of 3D building model from close-range photogrammetry is categorising in LoD3. Figure 12 shows the different LoD of building model between the base model and close-range photogrammetry.


Geoinformation Science Journal, Vol. 10, No. 1, 2010, pp: 49-60

3D building from close-range photogrammetry


Block model

Figure 12 Different LOD of Building Model between The Base Model And Close-Range Photogrammetry The results between the two models are compared in term of geometry and textures. The comparison is shown in Table 1. Table 1 Comparison between block model and 3D building model from close range photogrammetry 3D model Block model 3D Building model

Geometry Consist of 8 edges No roof Consist of 114 edges. With roof

Textures No textures Textures

The block model is a simple 3D model and it is shown such as a box. Block model consist of 8 edges, without roof and without textures. The edges are defined from the 3D block model in base model. The 3D building model as compared to close-range photogrammetry, the building model is produced same as the building from real world. The building model consists of 114 edges including the roof edges and the building model has textures. Therefore, the data from close-range photogrammetry is needed in order to develop photorealistic 3D city model.


PhotoModeler Accuracy Assessment on 3D Building

The accuracy of photogrammetric measurement are depends on camera resolution, quality of camera calibration, geometry of the camera position, and the precision of marking location on the images (PhotoModeler, 2009). High camera resolution with good quality camera calibration is helps to achieve high accuracies in precision marking location on the images. This is the reason why the high camera calibration with good calibration results is used in this project. The geometry of the camera position refers to angles between the camera stations. The low angle of the camera station gives low accuracy of the projects. The precision of marking location on the images shows the quality of measurement. Precision is determines by the number of residual value which the lower residual value shows the good marking precision. The geometry of the camera position and marks precision location is shown on Table 2.


Geoinformation Science Journal, Vol. 10, No. 1, 2010, pp: 49-60

Table 2 Accuracy Assessment of point location. Id




1 2 42 75 76

627182.6509 627186.0713 627180.9008 627179.0669 627178.9186

172769.4714 172769.3138 172778.7513 172776.5189 172771.7541

2.1349 0.0150 0.0491 2.2998 2.2551

X Precision 0.0001 0.0001 0.0001 0.0001 0.0001

Y Precision 0.0001 0.0001 0.0001 0.0001 0.0001

Z Precision 0.0001 0.0001 0.0001 0.0001 0.0001



0.0003 0.0013 0.0036 0.0049 0.0077

64.1656 81.7586 88.8676 89.4690 86.2181

In Table 2, Id refers to the number identifying of marking points, XYZ value refers to the marking point coordinates, XYZ Precision shows marking point precision, RMS is refers to the Root Mean Square which shows the average marking precision of the point in pixels, and last column is shows the point angles between the photo. The overall RMS value of the building model that has been developed is 0.005 pixels and according to PhotoModeler (2009) the RMS value should be less than 3 pixels in order to have good photogrammetric project.



As a conclusion, the CRP method has capabilities to solve the constraint of aerial images in developing 3D model for small building in 3D city model development. In addition, the combination of aerial and close-range photogrammetric can be achieved into LOD3 which consist of details building geometry and textures.

REFERENCES Cooper, M.A.R., and Robson, S., (2000). Theory of Close-Range Photogrammetry. In Atkinson, K.B. (Ed.) Close Range Photogrammetry and Machine Vision (pp 9–50). Pennsylvania State University : Whittles Publishing. Dallas, R.W.A., (1996) Architectural and Archaeological Photogrammetry. In Atkinson, K.B. (Ed.) Close Range Photogrammetry and Machine Vision (pp283). Pennsylvania State University : Whittles Publishing. ESRI (2008), The Multipatch Geometry Type. Retrieved December 2010, from Flamanc, D., Maillet, G., and Jibrini, H. (2003). 3D city models : An operational approach using aerial images and cadastral maps. In Ebner, H., Heipke, C., Mayer, H., and Pakzad, K., editors, Photogrammetric Image Analysis, pages 53.58, Munich, Germany. ISPRS Hallert B. (1960). Photogrammetry, Basic Principles and General Survey. McGraw-Hill book company, Inc., New York. Jurgen, D., Kolbe, T.H., Falko, L., Takis, S., Karin, T. (2006) The Virtual 3D City Model Of Berlin. University of Potsdam, Germany. Kobayashi, L (2007) Photogrammetry and 3D city modeling. Arizona State University, USA. PhotoModeler (2009). Photomodeler Pro6.0 Help. Retrieved June 2009, 58

Geoinformation Science Journal, Vol. 10, No. 1, 2010, pp: 49-60

from Vermiej, M., and Zlatanova, S. (2001). Semi-automatic 3D building reconstruction using Softplotter. Delft University of Technology, Department of Geodesy, The Netherlands. Vesna, S. (2008). Terrestrial photogrammetry and application to modeling architectural objects. Department of Architecture, Faculty of Technical Sciences, University of Novi Sad, Serbia. Hanke, K. and Ebrahim, M. A.B., (1997). A low cost 3D-measurement tool for architectural and archaeological applications. International Archives of Photogrammetry and Remote Sensing, 32(5C1B): 113–120


Nor ‘ Ainah Amat is M.Sc student at Faculty of Geoinformation and Real Estate, Universiti Teknologi Malaysia. She hold her BSc (Geomatic) degree from Universiti Teknologi Malaysia. She is member of Photogrammetry and Laser Scanning Research Group and her current M.Sc research project are focus on 3D city development using Photogrammetry approach.

Dr. Halim Bin Setan is a professor at the Faculty of Geoinformation Science and Engineering, Universiti Teknologi Malaysia. He holds B.Sc. (Hons.) in Surveying and Mapping Sciences from North East London Polytechnic (England), M.Sc. in Geodetic Science from Ohio State University (USA) and Ph.D from City University, London (England). His current research interests focus on precise 3D measurement, deformation monitoring, least squares estimation and 3D modeling.


Geoinformation Science Journal, Vol. 10, No. 1, 2010, pp: 49-60

Dr. Zulkepli Bin Majid is a senior lecturer at the Faculty of Geoinformation Science & Engineering, Universiti Teknologi Malaysia. Currently, he is the head of Photogrammetry and Laser Scanning Research Group. He holds B.Sc (Land Surveying) degree, M.Sc (Photogrammetry) degree and a PhD (Medical Photogrammetry) degree from Universiti Teknologi Malaysia, Malaysia. His research interests lie in the areas of Photogrammetry and Laser Scanning for various applications. His current research projects are the application of photogrammetry in forensic and assets data capture using terrestrial laser scanner.