Research Of Tracking Surgical Instrument Based On Binocular Stereo Vision In Surgical Navigation System

The application of binocular stereo vision in the field of surgery brings greatimprovement to surgical accuracy and success rates.Since the surgical instruments areactuators of the surgeries,tracking the surgical instruments, especially the end of operatingpoint of surgical instruments becomes one of the most critical technologies in the field ofsurgical navigation system.In order to achieve fast and accurate tracking of surgical instruments, this paperdesigned surgical instruments attached with marking boards, studied the calibrationtechniques of camera and projector,detection and matching techniques of marking points,and tracking technology based on color information.Specific contents and results of the research are as follows:1.Based on the study of current camera calibration techniques, this paper adopted animplicit camera calibration method based on BP neural network. This method builds directmapping from image coordinates to world coordinates by training the neural network, givesfull consideration to the various nonlinear factors ignored by conventional camera models,then improves the calibration accuracy of the camera.2. This paper designed surgical instruments attached with marking boards, applied sub-pixel corner point detection algorithm based on Gaussian surface(on the basis of classicHarris algorithm), and used Freeman chain to achieve quick matching of marking pointsbased on selecting algorithm.3. This paper established local coordinate system of surgical instruments, resulted inleast-squares solution of over-determined equation according to the continuous acquisitionsof rotation locus of the surgical instruments based on non-contact register method,completed the register of the working point.4. This paper studied Meanshift and Camshift tracking algorithms, subdivided Huecomponent by Saturation component, increased saturation information of back projection,solved the problem of bad robustness under similar background based on traditionalCamshift algorithms. This paper also studied the tracking strategy when marking board wasshortly blocked, estimated marking points at blocked moments by space linear least-quaresfitting method, avoided tracking failures caused by short blocking. Experiments showed the projection calibration and marking points detection achievedsub-pixel accuracy,the error of camera calibration,working points register and tracking wereless than0.8mm.Initial results achieved in this paper met the tracking requirements onsurgical instruments during surgery with great application prospect and practicalsignificance.