3D PHOTOGRAMMETRIC RECORDING USING DLT AND CAD Francesco Guerra 1 , Vassilios Tsioukas 2 1 Università IUAV di Venezia, Laboratorio di fotogrammetria, CIRCE, S.Croce 1624, I-30135 Venice, ITALY, guerra2@iuav.it 2 The Democritus University of Thrace, Dept. of Architectural Engineering, New Building of Central Library, Kimmeria, 67100, Xanthi GREECE, vtsiouka@arch.duth.gr Commission V, WG V/2 KEY WORDS: Close Range Photogrammetry, 3-D feature extraction, Calibration, Architectural Heritage, Digital Mapping ABSTRACT: Novice users currently use Digital Photogrammetric Applications increasingly in order to record close range objects of great cultural value or for the monitoring of sensitive objects or procedures (industrial and medical applications). Monoscopic procedures (digital rectification) have been embedded recently, in many popular commercial applications. Also, CAD and CAM applications are gaining even more their use by people of low knowledge background in Photogrammetry, namely Archaeologists, Medical Doctors, Mechanical Engineers, etc. 3D photogrammetric applications are very important and could be part of the every day activity by the above-mentioned professionals. Since the real world is 3D and the demand for accurate recording and representation is increasing even more with the development of new high tech computing systems and reproduction devices, the development of easy to use 3D photogrammtric software applications is more than ever a necessity. Ground Control Points (GCP) measurements, bundle adjustment of colinearity equations and camera calibration procedures applied to stereoscoping and convergent photogrammetric images offer the most efficient and accurate solution. However, the above-mentioned tasks are hard to be applied by many potential users with low implication in photogrammetric projects. There are also some cases where these tasks cannot be applied due to the uncertainty or short of information about basic camera calibration parameters. In these cases, Direct Linear Transformation offers a way to extract reliable 3D information with just the use of some GCPs which could be easily measured using modern conventional reflectorless Total Stations. This paper is describing our attempt to provide a low cost CAD software application embedding the DLT photogrammetric processing to create accurate 3D drawings of close range objects. The use and distribution of the application is free for academics while the demands in hardware configuration (CPU and graphics card) are very low. 1. INTRODUCTION The use of DLT (Direct Linear Transformation) in order to calculate the object space of imaged objects has given very good results in the past (Douglas, et.al. 2002, Savapol and Armenakis 1998, Remondino, 2002). It is a very well documented process that has the ability to provide, in a sort time and with adequate accuracy the 3D information of imaged details on an object ignoring the physical mathematic model of the central projection. It embeds parameters to model not only the extrinsic but also the intrinsic camera values. Even most important is the use of parameters that describe the most significant camera calibration parameters such as the radial distortion and the camera constant value, in the DLT equation model (Tsai, 1987, Dermanis, 1991). The sensor to be used in the DLT process can be any off-the- shelf imaging device such as a typical dSLR camera with a camera lens of low radial distortion values or any low-cost (but of lens with high values in radial distortion) compact digital camera. In a typical bundle adjustment procedure using the colinearity equation model, the imaging sensors should be only of one kind in order to provide the most accurate results. Even different camera constants of the same sensor might increase the processing complexity and decrease the overall precision. After all, in a typical photogrammetric project a single camera is used to provide the projects images. Thus, most of the commercial photogrammetric applications solve projects of images coming from a single camera. In the case of DLT there is no limitation of the types and number of sensors that will be used to record an object and the computation of parameters describing the intrinsic camera values is unique for every image. Consequently, higher accuracy is expected from the DLT solution in a typical Single Camera/Multi Image photogrammetric project. 2. MATHEMATIC MODEL The DLT equations model consists of the following unknowns and observations 0 1 0 ) , , , , , , , , , , ( 11 10 9 4 3 2 1 11 10 9 4 3 2 1 = + ⋅ + ⋅ + ⋅ + ⋅ + ⋅ + ⋅ − ⇔ = Z L Y L X L L Z L Y L X L x L L L L L L L Z Y X x f (1) 0 1 0 ) , , , , , , , , , , ( 11 10 9 8 7 6 5 11 10 9 8 7 6 5 = + ⋅ + ⋅ + ⋅ + ⋅ + ⋅ + ⋅ − ⇔ = Z L Y L X L L Z L Y L X L y L L L L L L L Z Y X y g (1) where x, y are the image coordinates of the control and tie points on the images and X, Y, Z are their 3D coordinates. The unknown parameters are the L1, .. L11 values. The unknown 321