3D Imaging And Object Reconstruction From Raster Projection

As the fast development of modern industry society, the technologies of three-dimensional object modeling and measuring play more and more important roles. At the same time, the practical needs put forward higher and higher requirements on the technologies. In this thesis, a system for raster projection 3D imaging and 3D object reconstruction and measuring is proposed, combining the theories and techniques of geometric projection, computer graphics and virtual reality. This system has the merits of simple structures, good precision, fast speed, wide applications, high automation, etc.In the system, a set of binary codes are used to acquire the approximate phases of points in a distortional raster projection image firstly. Then, a set of phase-shift codes are used to acquire the accurate phases of the points. Because the phase unwrapping in the system is implemented inside the binary code strips, the accuracy of phase unwrapping has been promoted.In the raster projection image processing, the author of this thesis proposes a few optimized methods. Because the raster projection image is usually degraded by random interference from such as environment, illumination, camera lens, image scanning, quantization, transmission, etc., a method of fast background correction is proposed for improving the image quality and contrast. In addition, a fast recursive median filter and algorithm is developed for image filtering and image enhancement. Since the raster projection image consists of black and white strips and has highercontrast than non-raster projection image, an adaptive thresholding algorithm is proposed for automatic image thresholding. Considering the segmentation would produce some isolated noisy spots, a method, named binary image rotating filter method, is adopted to remove the spots. Furthermore, the image pixels are classified into two classes: invalid pixels and valid pixels, and only the valid pixels are processed. As the result, the image processing speed and precision are promoted.In object measuring, the possible errors in 3D object reconstruction and measuring are analyzed. A new method is proposed for error correction and the accuracy of measuring is improved. The image registration is also analyzed and an effective method of registration is presented as well for calibration and big object measuring.In 3D object reconstruction, three methods are utilized: volume reconstruction, surface reconstruction and projection reconstruction. In surface reconstruction, a fast algorithm of VRML surface reconstruction is proposed. For 3D visualization of the reconstructed objects, a VR viewer is developed using OpenGL technology. Especially, a series of methods to enhance the display effect, such as smoothening, transparentization, shadowing, illuminating, material rendering, coloring and so on, are developed. To improve the convenience of visualization, a few functions, such as dragging, free moving, translation, etc., are developed as well.