Field Calibration Method Of Stereo Vision Sensor And Implementation

Stereo vision measurement technology is the one of commonly used implementation methods of vision measurement, by simulating the way that human visual system processes image information of three-dimensional scene and using the principle of optical triangulation to calculate the 3-D coordinate of object, finally obtaining geometrical size measurement and position of object. Stereo vision calibration is the prerequisite for precise measurements, only by calibrating can the stereo vision system perform measurement tasks, and the measurement precision is closely related to the calibration precision of the stereo vision system. Therefore, to achieve high-precision calibration of stereo vision is important significance.This thesis mainly focuses on high-precision calibration of stereo vision system, and according to the actual application needs, we design an overall calibration scheme of stereo vision system and implement the system calibration based on C++language and OpenCV library. First of all, this thesis analyzes the basic principle of stereoscopic measurements system to determine the composition of stereo vision system, in the mean time we analyze the camera imaging process that a single three-dimensional point transforms to the image point on image plane in camera, to establish the camera model and stereo vision sensor system measurement model. Then according to the measurement accuracy requirement, we determine the calibration algorithm that using a planar target, and considering the stereo vision system calibration process, we designed an overall calibration scheme of stereo vision system, while designed a target which suited for stereo vision system calibration. According to the geometric properties of feature points on the planar target image, we researched algorithm which achieved the goal of detecting elliptical feature points on images automatically, and segmented every elliptical region separately to detect the sub-pixel edge of the ellipse, and obtained the image coordinate of center of ellipse using ellipse fitting method to calculate the analytical expression of ellipse. To meet the requirement of measurement accuracy, we expanded the feature points and obtained a lot of derived image coordinate of feature points. Finally, we set up a stereo vision measurement system, and calibrated the stereo system based on C++ and OpenCV library, and conducted error analysis.The experimental results suggest that the calibrated process raised in this thesis can calibrate the stereo vision system with high-precision and obtain the three-dimensional coordinate of body joints in the key frame image.