Robot Vision Experiment Platform And Its Algorithm Design

Binocular stereo vision has a wide application prospect. With the development of optics, electronics and computer technology, it has been more and more practical and will be used in many fields in life and production. In the early 1980s, Marr's theory of computational vision built an explicit system for the investigation of computer vision. It is a potent impetus of the machine vision research. But the bottleneck of the computer vision is still the question of correspondence, that is finding the spots corresponding to the same three-dimensional point in the two camera pictures. A research of obtaining deep information, using a laser token as the images matching feature, is discussed in this paper.The background of machine vision and the bionic base of binocular vision are introduced and the principium of binocular stereo vision is analyzed at the first of this paper. Every degree of freedom of the 4-DOF binocular stereo vision mechanism system can be controlled individually, so transformation formula of each reference frame must be presented clearly. The relationship of every reference frame in the process of transformation from pixel matrix reference frame to world reference frame is analyzed. The algorithm of obtain three-dimensional coordinates of a special point according to the corresponding coordinates in the pixel matrix reference frame is discussed as well as camera calibration method that use least square method to calculate the parameter.The mechanism design scheme and its dynamics character are discussed. The design aim of 4-DOF binocular stereo vision system and the configuration characteristic are analyzed. The dynamics characteristic of the mechanism is analyzed using Lagrange method according to the 4-DOF mechanism, and the dynamics model is established. The dynamics simulation is carried through, and the response time that the mechanism start from static state while orientate a target is analyzed. This thesis discussed a new method of images matching. Using a laser token as a feature can simplify the matching process and reduce the calculating quantum. In an image with laser dot, the laser dot pixels'gray values are far larger than the gray values of the background. But the number of the laser dot pixels is much less than the background pixels number. The method of effectively using the laser dot pixels'lighteness is discussed. And the methods of how to recognize the laser dot and obtain the center of the laser dot are discussed as well.An experiment is carried through to obtain the camera parameters and test the precision of the system. A printed plane target object is adopted in the experiment. The coordinates of the targets are obtained with the 3-dimension-measuring instrument. Then, the unknown camera parameters can be worked out. With the camera parameters and coordinates of the targets in the pixel reference frame, the targets'3-dimensional coordinates in the world reference frame are figured out. The causation of the error between the computed and measured value is analyzed.All the work of this thesis shows an expecting achievement and builds the foundation for the farther research. With the development of the work, the binocular stereo vision system with laser token will acquire a satisfying performance.