The Key Technologies Research Of Humanoid Robot Localization Based On Binocular Stereo Vision

As a new field based on various interdisciplinary, the theory research and practical application of the computer vision has made great progress. Especially in recent years, the development of image processing technology and high performance of computer signal processing technology provide favorable conditions for computer vision research and application. Robots based on machine vision have been widely applied, Computer vision played an important role in location of object. Binocular stereo vision as important branch of computer vision is one of keys of computer vision research. This paper studied the application of stereo vision system to location of human face.The 3D reconstruction system includes three components: camera calibration, stereo correspondence and 3D information calculation. In camera calibrate, the method in this paper analyzes and compares common camera model and its calibration, bring forwards a new calibration method, namely the least two multiplication. During calibration, it can reconstruct the object through the projective matrix, avoid solving the parameter of the camera inside and outside, decrease calculation complexity. In stereo correspondence, the algorithm based on the centroids feature were extracted. The centroids of two images were directly utilized to matching for estimating the approximate position of human face. The algorithm is simple, easily realized, saving time. However, there are many false pairs. At the end, in order to resolving this problem the region match based on standard spot restraint is used to eliminate the false pairs and find new pairs at the same time. After the camera calibration and the projection of special points in two images are gained, the depth of the point can easily be calculated. In three-dimensional system, comparative position of two cameras should be obtained through calibration. The method in this paper avoids this complicated calculation. The experiment result indicates that the principle of the method is simple, the quantity of calculation is small, distance measure can come true.In this paper, a camera calibration method based on the least two multiplication, a matching method based on centroids and region, and a method to get 3D information are advanced respectively to three parts of 3D reconstruction system. These methods are all proved correct and feasible by experiments. As a result, a software prototype for 3D reconstruction is implemented by C++. The prototype realizes the purpose that a point of human face is reconstructed, and provides foundation for more far-ranging functions of stereo vision on robot.