Research On 3D Positioning Technique Based On Binocular Stereo Vision

Stereo Vision is used to locate spatial point, which is an important research field of Computer Vision. Binocular Stereo Vision is the famous one in Stereo Vision, due to its characteristic of simulating the manner of man's eyes processing information, which has great applicability. Research is done on Binocular Stereo Vision's key technique, based on the project named on 3D Probability Model Assisted Prostate Tumor Biopsy Navigation System. The system can calculate the position of surgery instrument and instruct doctor locating the part of focus.The main technique of Binocular Stereo Vision is composed of camera calibration, stereo matching and 3D coordinates computing. Camera calibration is a key one. For the Surgical Navigation System of Computer-Assisted Surgery must have simple operation, high precision and good stability, the method of camera calibration must be simple, flexible and fast. We adopt Zhang's flexible new camera calibration method, and solve the problem about how to automatically get the pixel coordinates and world coordinates of all the characteristic points on the calibration board by homography matrix. Besides, we consider the effect of tangential distortion to improve the Zhang's shortage only calibrating radial distortion, which has better result and high precision. According to the actual need, we put forward a stereo matching method based on characteristics points, which can obtain sub pixel precision. In the process of matching, we propose a new way to filter points based on the jump of gray, in the meantime we utilize polar line restrict to lower 2D research to 1D,which consumedly reduce the time and advance the efficiency. At the last, the 3D coordinates are given by least square method through the results of stereo matching.The experiment has shown that the stereo positioning precision of the system is high and the error is small under the current experiment condition, also the operation is simple.