COMPUTER VISION : APPLICATION IN PHOTOGRAMMETRY

Abstract

Computer Vision is the ability of a computer system to perform certain tasks normally associated with human vision, including sensing, image formation, image analysis and image interpretation or decision making. It uses the knowledge of traditional engineering and computer programming for the different parts of a process. Computer Vision System is being used in Robotics, Optical Engineering, Signal and Image processing, Psychophysics, Neurophysiology for quite some times. Only recently it has found its application in photogrammetry as well and some photogrammetrists have designed systems based on Computer Vision System which are now popularly known as a Digital Photogrammetric System. The application of Computer Vision in the photogrammetry is to recognise the physical characteristics of the objects and to recreate its 3-D geometry. The 3-D geometry may be used for preparing a topographic map. Many developments and improvements in remote sensing techniques have been directly related to the outstanding advances in microcomputers, graphics boards and high resolution monitor technology because it deals directly with the digital data. In contrast to the above, in conventional photogrammetric approach, hard copy form of photographs are processed using standard stereoplotters, either analog or analytical. Even though the rate of development in photogrammetry in recent years has been low, its contribution and importance continue to be major, because of the importance of the three dimensional information that it provides. As both disciplines are closely related, both dealing with imagery and deriving spatial information from it, it seems that_waym,should-be-found-out-to-Close-the-gatib-etT-/-6-en---Eh-eM- and to put together the specific advantages of each discipline. This directly leads to the concept of Computer Vision, a system that shows the potential of doing the job of a stereoplotter. In this,the stereo images in digital form, are directly fed into and processed by a computer.pThis approach offers new, lower cost alternatives to stereoplotters. In this dissertation, an attempt is made to develope such a computer vision system, which simulate a photogrammetric plotting system. The potential, characteristics and operational aspects of the system are explained. As a test case, swiss photo-pair at the average scale of 1:4300 with focal length of 153.18mm was taken and digitised on HP scanner with 300DPI in 256 grey level. DLT approach was used to calculate object space coordinates. The results obtained were compared with that of BC2 analytical plotter. Vector file is overlaid on the displayed image to generate a digital map in the form of DTM. The accuracy obtained by this system, developed in the present study, is comparable with any II order plotter.