Study On Monocular 3D Vision For Ureteroscopy

Currently,3D endoscopes have become a hot spot of endoscope research.However,there are still three problems in the current 3D endoscope research: 1.The current 3D endoscope is expensive,and that makes it difficult to be widely promoted in the hospital.2.The current implementations of 3D endoscope are mostly based on the binocular vision system,so it cannot match the monocular endoscope system widely used in hospitals.3.It is very difficult to apply the binocular vision system in the cation of microscopic endoscopes such as ureteropelvic endoscopy.This paper proposes a monocular-to-dual-purpose observation system based on Structure from Motion(SFM)technology combined Stereo Matching to achieve 3D vision with monocular endoscope.The main work of this paper is as follows:1.For the endoscopic camera parameters,this paper uses the pinhole model,employs the Chessboard calibration method of Professor Zhang Zhengyou to solve the camera’s internal parameters and distortion parameters.2.The motion solution of the endoscope is the basic work of the reconstruction of the perspective proposed by this paper.In this paper,the scale-invariant feature transform(SIFT)matching algorithm which is more robust,is used to solve the correspondence between the pixels in different positions of the camera,and the solution of Basic matrix and 3D reconstruction work helps obtain the posture information of the endoscope.3.In terms of stereo matching,this paper uses 3D Label Stereo Matching method using Local Expansion Moves to obtain dense matching disparity data.4.After obtaining the posture information of the endoscope and the dense matching parallax data,this article uses the 3D reconstruction technology to restore the three-dimensional information of the field of view,and uses the principle of binocular vision to re-project the right-eye view.And the final experiment shows that the algorithm,which provides a basis for the realization of stereoscopic vision system for the monocular ureteroscopy,is feasible and can solve the virtual right eye view.