Research On Key Techniques Of Surgical Navigation Based On Mixed Reality

In recent years,with the development of computer vision,digital simulation and mixed reality technologies as well as the miniaturization of the related equipments,mixed reality technology(Mixed Reality,MR)has made great progress in the visualization and the navigation of medical surgery.Among them,the virtual-real registration technology is an important foundation for implementing a surgical navigation system based on mixed reality.Many experts,scholars and organizations at home and abroad have conducted research on virtual and real registration algorithms.The background of this subject is to provide surgical navigation and training for doctors during rescue operations on a helicopter.This scene is unstable and uncontrollable compared with conventional laboratory or hospital scenes.Therefore,in this scenario,the virtual-real registration algorithm must come over the high dynamics and have higher robustness.In this regard,the research works of this dissertation mainly covers the following aspects:First of all,the detection and matching of target features is the first step of the algorithm,and its accuracy and matching time are of great significance to the accuracy and real-time performance of the system.This paper proposes a feature matching algorithm based on feature database.In this algorithm,the existing digital spine model is first used for feature extraction and establishing a feature library.In the experiment,the speeded up robust features(SURF))algorithm and the random sampling consensus algorithm(Random Sample Consensus,RANSAC)are used to detect and describe the features of the image returned by the camera in real time,and compare with this feature library.Experiments show that our improved algorithm has strong robustness in terms of illumination changes,partial occlusion,motion blur,scale,viewing angle,and rotation changes.Secondly,for the highly dynamic scene of rescue on the helicopter,an Extended Kalman Filter(EKF)algorithm is added to the virtual-real registration algorithm to predict and track the target state.In addition,the Extended Kalman filter algorithm is modified,and the tracking coordinates of the filter are calculated through the clustering method,which greatly improves the stability of target tracking.The addition of the target tracking algorithm also improves the robustness of the virtual-real registration algorithm under high dynamic conditions.Finally,the improved iterative closest point algorithm(ICP)is used for real-time registration.Experiments show that the algorithm framework proposed in this dissertation can solve the virtual-real registration problem in a high dynamic environment.