Tracking And Positioning Of Surgical Instrument Based On Binocular Stereo Vision

Binocular stereo vision is used to get the three dimensional information of the scene based on the image geometry theory requiring two cameras to observe the identical scenery between two different visual angle. The techniques in three dimensional reconstruction and object tracking based on the theory of binocular stereo vision can be used in Surgical Navigation System of Computer-Assisted Surgery to instruct surgery doctor completing the high difficult operations, which have very important significance in improving operative location and lessening operative injury.The positioning of surgical instrument based on binocular vision focuses on the research of camera calibration and stereo matching. Considering that the surgical navigation system requires convenient operation,high-precision location and well stability, Zhang's plane template two-step method is adopted to flexibly and quickly get the exterior and inner parameters in camera calibration. In order to improve the calibration precision, tangential distortion adjustment is added into the camera imaging model. For stereo matching of dual images, a modified compact algorithm is adopted on the epipolar rectification. And after rectification of stereo pairs, stereo matching based on corner feature is done by simultaneously extracting corners and searching for the corresponding points nearby the horizontal scan lines of candidate points. By this method, it will consumedly reduce the search range and time, which can reach real-time matching. In the study of surgical instrument tracking, the Camshift algorithm combined with Kalman filter is proposed. Kalman filter is used to forecast possible position of target, then Camshift search the real position neighbourhood of the possible position. Meanwhile, utilizing the distance of two color histograms to judge the case of serious occlusion or disturbances of track target, which, to some extern, efficiently solve the tracking problem that the target is completed occlusion for a little long time. This algorithm is robust that can well satisfy kinds of special cases in target tracking.The experiment has shown that system can fulfill real-time tracking and positioning and get the precision moving path of object under the current experiment condition.