Research On Key Technologies Of Endoscopic Three-Dimensional Imaging

In recent years,minimally invasive surgery has been rapidly developed,and the endoscopic three-dimensional imaging technology to assist minimally invasive surgery has also been favored by researchers.In this paper,the key technologies of 3D endoscopic imaging are studied deeply,and an ORB-SLAM2 system based on Semi-Global Matching algorithm(SGM)is proposed to realize the 3D imaging of the endoscope.The main contents of this paper are as follows.(1)The ideal imaging model of camera is introduced,the calibration methods of monocular camera and binocular camera are derived,and the stereoscopic correction method based on polar line correction is studied.Two calibration methods are programmed,and the correctness of the algorithm is verified by experiments using standard calibration set.Finally,the endoscope used in this paper is calibrated.(2)We study the implementation principles of SIFT,SURF,BRISK and ORB for feature extraction,and evaluate the performance of these algorithms from three aspects of feature extraction time,rotation transformation and scale transformation.The experiment shows that the feature extraction time of ORB feature extraction algorithm is shorter and the feature matching accuracy is higher.(3)The principles of SGM and AD-Census stereo matching algorithms are studied and programmed.The performance of the two algorithms is compared by using the standard image set.The results show that although the matching accuracy of SGM is slightly poor,its fast matching speed is more suitable for real-time 3D reconstruction.(4)An ORB-SLAM2 3D imaging system based on SGM is proposed,and GPU and multithreading are used to improve the real-time performance of the system.The results of 3D reconstruction of plaster model and human gastrointestinal data set were quantitatively analyzed from the aspects of accuracy estimation,point cloud surface fitting,3D point cloud number,reconstruction accuracy and so on.It was proved that the performance of ORB-SLAM2 3D imaging system based on SGM was superior to the original ORB-SLAM2 system in time complexity,accuracy and other aspects.