Research On 3D Reconstruction Method Of Stomach Working Environment For Magnetorheological Medical Capsule Robot

According to the 2022 World Cancer Report released by the World Health Organization,the number of residents in China suffering from colorectal cancer and gastric cancer ranks second and third respectively.In order to detect,treat,and improve the cure rate of gastrointestinal diseases early,physical examination and screening are crucial.Medical capsule robots are an emerging means for the examination of human gastrointestinal diseases,which is of great significance for the development of modern medical fields.However,the images captured by medical capsule robots during movement have narrow field of view,complex changes in inclination and rotation angles,and unknown overlapping areas,resulting in the inability to complete threedimensional reconstruction of the gastric environment.Aiming at the current problems in the visual diagnosis of medical capsule robots,this article has conducted research in the following aspects.(1)The overall framework of 3D reconstruction system for the stomach working environment of a magnetorheological medical capsule robot is designed.The structure and working principle of the capsule robot are studied,the fold characteristics in the stomach,various internal characteristics and layered image characteristics are analyzed,the camera coordinate system and world coordinate system centered on the capsule robot are established,and the transformation relationships among the four coordinate systems are deduced.According to the requirement of three-dimensional reconstruction of working environment of stomach,the software module requirement design and capsule camera selection are carried out.(2)The methods of image feature detection and matching for stomach working environment are proposed.The bilateral filtering algorithm and MSRCR algorithm are used to filter and enhance the image respectively.A stomach image feature extraction algorithm based on improved SIFT algorithm is proposed,which optimizes the feature vector descriptor to 96 dimensions and reduces the time of algorithm feature detection;The fast nearest neighbor search algorithm is used to match the feature points,and the image local feature topological constraint method is introduced to perform the image feature precise matching.Simulation experiments are carried out to verify the feasibility of the proposed stomach image feature matching algorithm.(3)A 3D reconstruction method for the working environment of the stomach of a magnetorheological medical capsule robot is proposed.The EPn P method is used to solve the camera pose parameters,and the beam adjustment method based on LM algorithm is introduced to reduce the re-projection error,achieving incremental SFM algorithm for 3D sparse model reconstruction;The probability models of camera cluster Ci and surface triangle t are calculated using the general expectation maximization algorithm;Established a central pixel patch model,completed patch expansion and filtering,and achieved dense reconstruction of the stomach model;A simulation experiment is designed to verify the superiority of the 3D reconstruction method.(4)An experimental study on 3D reconstruction system for stomach working environment of the magnetorheological medical capsule robot is carried out.The experimental platform of reconstruction system was built,and the software platform of the system is designed and developed according to 3D reconstruction requirements of stomach working environment.A 3D reconstruction experiment on a gastric model and a 3D reconstruction test on a pig stomach were conducted to verify 3D reconstruction effect of this system.The thesis has 75 diagrams,15 tables,and 111 references.