The Research Of Image Registration Of Robot Assisted Surgical Navigation System

The positive image of the focus, not only clearly visible on the vision, but also convenientfor doctors to operate when the operation,is essential for the surgeon in robotic-assisted surgery,Therefore, how to retrieve positive image information of the focus has very important researchsignificance. To get the positive image, it is needed to registrate and then merge the image bymulti-angle of view. Therefore,we propose registration based on sequence images (images takenfrom different angles) and this paper studies the sequence of image registration based on externalmarkers and the sequence of image registration based on feature points.In this paper, the main research contents and the results are as follows:1.The design of the system model of image registration algorithm (including the hardwaresystem and software process). The image registration step is designed according to the imagecharacteristics; Comparised and analysed the space transformation model of image registration,and concrete example is given for each kind of transformation and using the environment;Contrast analysis of the advantages and disadvantages of three registration method, proposingthe sequence of image registration based on markers and image registration based on featurepoint sequence.2. Asequence of image registration based on external markers is proposed in this paper.Theblue marked points that is easy to identify and extract and adaptive dynamic window is designedFirstly,mark point area is extracted and calculate the coordinates of the marker centroid, then useimage matching method based on Laplacian matrix registrate the two different number of pointsset according to the sequence of the image registration strategy. Finally, images registration areweighted fusion, in order to obtain head image of positive image.3.The natural features of the head image is used to study the sequence image registrationSURF algorithm (group number,layer number,Hessian matrix determinant threshold) has realizedreal-time, accuracy and robustness of the registration algorithm. First feature points of imagesequences are extracted by using SURF algorithm,feature points are coarse matched according tothe Euclidean distance of feature points, then using RANSAC algorithm for feature pointmatching, improve the accuracy of registration, according to the sequence of image registrationstrategy with the camera around the final image, then refining registration feature points by usingRANSAC algorithm to improve the accuracy of registration according to the final imageacquired by sequence of image registration strategy in this paper and revising the picturesacquired by using Sine transform.Finally it can get head image of positive image by using the acquired by using Sine transform.Finally it can get head image of positive image by using theweighted image fusion.Experimental results show that the proposed sequence image registration method can verygood get lesions positive image information.The fusion image and positive face image of rootmean square error and the average error based on coupled with markers in the sequence of imageregistration and the sequence of image registration based on feature points finally the fusionimage and positive face image are smaller, conform to the requirements of the clinicalapplication.