Positioning, Based On Stereo Vision For Mobile Robot Target

With mobile robot being applied to military area, manufacturing and daily llife etc, it is an active and promising research area and attracts much more attention. Mobile robot's localization and navigation based on stereo vision is becoming one of important research direction of intelligent robot. In this thesis, the problem of camera calbiration,object feature extraction and feature matching , as well as object localization have been investigated in details based on AS-RF research platform, These algorithm are implemented by VC++6.0 and OpenCV function library, the main contents studied are as follow:Firstly, cameras's intrinsic parameters and extrinsic breakway cabliration method are presented in the thesis, it is the way between traditional calibration and self-calibration. Intrinsic parameters are calibrated by Zhengyou Zhang's way, calibrationg template comers' image coordinate are obtained by handwork, afterward get precise intrinsic parameters by experiment and calculating. Besides, extrinsic parameters's calibration is readily computed by left camera and right cameras' relationship and choosing left camera as reference model. The calibration has the merits of high precision, easily operation and fast speed.Secondly, the stereo matching way based on feature has been studied, takes the preprocess method fit for the task. In image segmentation, this paper proposes self-adaptive threshold method which has been changed a bit to make it more suitable for the task. In feature extracting, because of affect of shadow, and object has circle's character, therefore Hough transformation is applied to accurately extract object's centre of the circle. Then feature matching of left image's object and right image' object under the constraint of epipolar line.Finally , combing the camera's parameters and feature of object, the 3D coordinate of object' feature can be obtained easily by inverse transform of camera imaging, then object's position is known.The results experiment show that proposed way has high precision, which will provide a good basis for future research on robot navigation.