Research On Intelligent Surgery Aided Positioning System Based On Binocular Vision

Intelligent surgical assisted positioning systems have developed rapidly in recent years,and the real-time and accuracy of surgical operations have been improved.Among them,the optical positioning system is a key part of intelligent surgical navigation.It uses the spatial position sensor information to determine the spatial position of the surgical marker point,and tracks the surgical instrument and the surgical focus point to obtain the relative positional relationship between them.The research of optical positioning in intelligent surgery assisted positioning system is also one of the current hot research fields.In the context of surgical assisted positioning,a binocular stereo vision system is established in this paper,and the binocular camera is calibrated using the Zhang Zhengyou calibration method of the checkerboard calibration board;after the camera calibration,the improved stereo matching algorithm is used to get the parallax value;then based on binocular stereo vision to identify and track surgical instruments to improve the success rate of surgical operations.In order to achieve this goal,this article has done the following work.First,a simple configuration of binocular stereo vision positioning system was built,which included: MYNT camera D1000 version,Dell computer,binocular camera calibration template,surgical instrument positioning marker board.Second,the calibration method used in this article is the Zhang Zhengyou calibration method.It compares the four commonly used calibration boards in this method,and chooses to use a checkerboard calibration to calibrate the binocular camera.The calibration board can detect corners more easily and clearly than checkerboard which is commonly used,under the conditions of noise and uneven illumination.After calibration,the internal parameters,external parameters,rotation matrix,translation matrix and distortion parameters of the binocular camera are more accurate.After calibration,the internal parameters,external parameters,rotation matrix,translation matrix and distortion parameters of the binocular camera are more accurate.Third,this paper proposes to improve the traditional Census transform in the initial Cost Calculation stage of stereo matching,and add gradient information and color information for calculation;in the Cost Aggregation stage,introduce propagation filtering,and then perform parallax value calculation and optimization.When the algorithm has an image dense training and sparse training result coefficient threshold value of 1 in all regions,the percentages of error pixels are 26.2% and 19.1%,respectively,which are at least 2.9% and 3.9% higher than the MANE,FC-DNN,and Ga Net REE＿RVC algorithms.And using standard images commonly used in Middlebury v3.0 to verify the experimental results of Adirondack,Motorcycle,Piano and Teddy,the mismatch rate of parallax under intensive training is 1.16%,which is at least 0.39% longer than the other three algorithms.It is 0.46 s,which is an increase of at least 0.5s;under sparse training,the parallax mismatch rate is 3.67%,which is an increase of at least 0.34%,and the duration is 0.18 s,the above data proves that the improved method in this paper improves the matching accuracy and has good robustness.Fourth,this article proposes a stereo vision-based calibration method for surgical instruments during the identification and tracking of surgical instruments.Using Ar Uco codes as marking points,the identification and tracking of single and multiple marking points can be realized.The method is within the required error range is proved by experiments.