Research On Real-time Three-dimensional Reconstruction Technology Based On Structured Light

In the field of computer vision, 3D(three-dimensional) reconstruction algorithm is the key research direction. With this technology, we can reconstruct the 3D model of an object that exists in reality. It is widely used in different areas, such as health care, three-dimensional measurement, robot navigation and heritage conservation. At present, there are many methods to reconstruct the 3D model of the object, but some of these methods rely on expensive and sophisticated equipment, or some have low reconstruction efficiency and requires much computing time, or some have poor reconstruction accuracy. Therefore, making a cost-effective and real-time 3D reconstruction system is significant, which is the research direction of this paper. In this thesis, I do research in real-time 3D Reconstruction algorithm based on structure light and its implementation, including the following areas:Equipment calibration. I proposed a system calibration method to locate handheld device by the assistant camera. We design a cube as a calibration reference in 3D-reconstruction handheld devices, and each surface on cube is attached on a different color checkerboard. Through the assistant camera shooting the cube, we can calculates he position of the cube by some method and then the handheld device are positioned. This method can be used to point cloud registration, thus the process speed of the system will be lifted.when studying the cubic surface checkerboard corners detection on cubic surface, we developed a corner in checkerboard selection algorithm based on Harris operator. At first we discuss the technical principles of corner extraction by using Harris operator. Then, according to the center-symmetry of the partial images located to corners, we developed a corner selection algorithm, which can extract corners in checkerboard quickly and accurately. That is useful for locating the cube and can also be used to enhance corner detection in conventional calibration algorithm.The real-time 3D reconstruction based on structured light. We propose a study a structured light pattern based on pseudo-random coding sequence, which with the advantage of simple encoding and easy decoding. Then, we study the parallel processing to color texture reconstruction under the structured light proposed in this paper. This method eliminate the texture mapping step latter to speed up the pace of entire 3D reconstruction system.Design and implement a real-time 3D reconstruction system, including hardware building and software with detail modules. The hardware devices uses in the system are inexpensive equipment. the software system is easy to operating and with friendly interaction. After the experiment, the system can reconstruct rapidly three-dimensional model of the object.