Font Size: a A A

Research On 3D Reconstruction Method Of Binocular Endoscope Based On Stereo Vision

Posted on:2020-07-25Degree:MasterType:Thesis
Country:ChinaCandidate:Y S CaoFull Text:PDF
GTID:2492306497457114Subject:Electronic Science and Technology
Abstract/Summary:PDF Full Text Request
Compared with traditional open ones,minimally invasive surgery has the advantages of small wounds,less surgical complications and a great reduction of pain.However,in the course of operation,problems such as low visibility and flexibility of surgical instruments into the wound,can lead to loss of visual and tactile feedback at the patient’s lesion.The medical endoscope has a simple structure,small volume,and easy introduction of a small wound,which satisfies the requirements of medical devices for minimally invasive surgery.It guides surgeons to perform effective surgical operation by providing images of human organs.However,the traditional medical endoscope can only acquire two-dimensional scenes,lacking the depth perception of the surgical scene,and is prone to misjudgment.With the deepening of stereo vision technology research,the 3D endoscope system constructed by it has the advantages of simple structure,good equipment compatibility and mature algorithm research.Such an endoscope improves the safety of surgical operations by completing the 3D reconstruction and enriching the depth information of the surgical scene.Based on stereo vision and electromagnetic induction technology,this paper is going to designs a binocular endoscopy system for body cavity reconstruction.By constructing a software and hardware platform,the high-definition endoscopic image capture and image processing is realized;By analyzing the calibration model of binocular endoscope,the relevant parameters for 3D reconstruction are obtained.By combining the adaptive control kernel regression algorithm,improving the guided filtering stereo matching algorithm to enhance the accuracy of the disparity map;Based on the principle of stereo reconstruction,the 3D reconstruction of the body cavity scene is finally realized.Main achievements of this research include:(1)The construction of a hardware and software interaction platform.Improve the structure of the binocular endoscope for the problem that the traditional binocular endoscope cannot be positioned in real time.A six-degree-of-freedom magnetic sensor and an LED cold light source are integrated,which solves the problems of instrument positioning and uneven illumination of the scene.The software platform realizes realtime synchronous acquisition of the left and right video images through the multithread synchronization mechanism.This step provides support for the realization of system functions and improves the efficiency of subsequent image processing.(2)The calibration of the binocular endoscopes.The linear and nonlinear camera calibration model is analyzed,and the Zhang’s calibration method is selected to calculate parameter matrices.After analyzing the stereo calibration model of the binocular endoscope and estimating the positional relation between the lenses,the polar line correction of the image could thus be further completed,on the basis of which the accuracy of the disparity map could be improved.(3)Research on stereo matching algorithm based on guided filtering.The traditional guided filtering algorithm has poor matching effect in deep discontinuous regions.In this paper,a new weighted guided filtering algorithm is proposed.The algorithm uses the adaptive kernel regression algorithm to calculate the local weight matrix,which assigns more weight to the edge pixels and plays the role of edge protection.The iterative adaptive control kernel regression algorithm is used to interpolate and smooth the initial disparity map to improve the disparity map accuracy.Finally,the experiments conducted on Middlebury benchmark demonstrate that the proposed algorithm produces good performance in precision and edge-preserving.(4)Realization of 3D reconstruction of binocular endoscope.Based on the principle of stereoscopic reconstruction,the point cloud reconstruction and the threedimensional visualization methods are analyzed.For the case where the position change of the binocular endoscope is unknown in use,a coordinate transformation model between the magnetic emitter and the binocular endoscope is established based on electromagnetic induction technology.On this basis,the moving distance and reference coordinates of the target point are estimated,thereby improving the controllability and safety of the instrument during use.
Keywords/Search Tags:Binocular endoscope, Camera calibration, Stereo matching, Electromagnetic sensor, 3D reconstruction
PDF Full Text Request
Related items