Research On Three-dimensional Reconstruction Method Of Binocular Endoscope

With the progress and development of science and technology,endoscopic minimally invasive surgery and people's lives are being inseparable which is widely used in otorhinolaryngology,heart valve surgery,celiac disease detection and other fields.Compared with traditional surgery,minimally invasive surgery not only can restore the surgery in a short time,but also can reduce the wound area.Advanced surgical robot platform can make the minimally invasive technology more secure,more easily complete the operation,leading this research more popular around the world.This research has combined a variety of subject knowledge about image processing,pattern recognition,information transmission and processing,automatic control into medical imaging industry.It provides more secure and reliable 3D data for doctors to improve the success rate of surgery.Based on the principle of binocular vision,this paper focuses on the 3D reconstruction of binocular endoscope.The research covers camera calibration,hand-eye calibration,image preprocessing,and feature matching and 3D reconstruction.Firstly,the binocular endoscopic system was built according to the requirements and performance of medical endoscope.The design of the whole system is studied,including the selection of hardware modules and the design of software.The software design mainly includes two modules,the calibration module and the3D reconstruction module.Secondly,the calibration of the camera and the hand-eye calibration of the manipulator are realized.The principle of human vision in the binocular model,the transformation relation among the coordinate systems,and the model of the camera are analyzed.Then,the intrinsic matrix of the external matrix are obtained when the binocular cameras are calibrated in Zhang zheng-you method.At the same time,the stereo geometry of the binocular vision system is analyzed,and the hand-eye calibration is studied.Thirdly,pre-processing and image correction are performed to prepare the stereo matching.In the aspects of feature extraction and stereo matching,the method of local feature matching is selected,and the SURF algorithm is improved.Experimental results show that the improved SURF algorithm not only improves the accuracy of matching,but also greatly reduces the matching time,ensuring the real-time character of the algorithm and better performance.Finally,the internal and external parameters of the calibrated camera in the third chapter are combined with the logarithm of the correct matching points obtained in the fourth chapter.The 3D coordinates of the object are calculated by the 3D reconstruction algorithm.According to the coordinate transformation relation,the 3D point coordinate in the camera coordinate system is converted to the robot arm coordinate system.Through the texture mapping approach,the models with visual and realistic under different coordinate system is reconstructed.