Calibration And Design Of Multi-View Video Acquisition System In Stereo Display System

Endoscopic minimally invasive medical devices are an organic combination of clinical medicine and the high-tech disciplines such as bioengineering,optics,precision manufacturing,image processing,medical materials,optoelectronic and electromechanical information,which have developed rapidly in recent years.Display terminal is an important part of endoscope system.And binocular stereoscopic glasses have been applied in endoscope system at present.In recent years,as the next generation of stereoscopic display technology to replace binocular stereoscopic glasses,nakedness-eye stereoscopic display technology is becoming more and more mature.This technology has the advantages of no auxiliary device on the head and suitable for multi-person and multi-angle viewing,avoiding the inherent defects of the binocular stereoscopic glasses.However,a key problem is that the visual information obtained by the traditional binocular imaging structure can't provide sufficient and accurate data for each subviewpoint of the multi-view naked stereo display system,which leads to various problems such as unsmoothly link between viewpoints,incomplete edges and holes in the multi-view stereo display system,and can't meet the requirements of commercial applications.Therefore,we propose to study more video sensor arrays composed of image sensors as the front segment which captures video in endoscope system,and use multi-view video information to reconstruct the scene and target in threedimensional.Then we use the three-dimensional reconstruction model to reconstruct the projection of information in each view point,and then complete the stereo view synthesis and display of each view point data.In this way,the requirement of nakedness-eye stereoscopic display terminal for imaging data is solved.For the above purposes,this thesis focuses on two key issues: the simulation platform of endoscope imaging system and the calibration of the camera system.The main work and innovations of this thesis are as follows:1.The hardware scheme of the multi-view video real-time acquisition subsystem is designed,and the experimental platform is built.The real-time acquisition of the multi-view video real-time acquisition subsystem is realized.The experiment shows that the system has reached the synchronous acquisition rate of 91FPS@VGA for 8-view video channels,and has the characteristics of low cost,good portability,easy configuration and deployment on site.2.Using LCD as the template carrier,the method of image acquisition and processing in high-precision intrinsic parameter of calibration is studied.The common-mode noise such as ambient light is suppressed,and the optimal method of image acquisition and processing is determined.The calibration accuracy reaches 0.087 pixels.3.Research on fast dynamic calibration algorithm is carried out.The SURF algorithm is used to extract the feature points of the initial image pair quickly,and the incremental matching method is used to match the feature points of the multi-camera picture to obtain the complete set of the extrinsic parameter matrix of camera.At the same time,the bundle adjustment iterative optimization is used to minimize the re-projection error.Finally,the multi-view epipolar correction method is used to obtain ideal multi-view image sequence output.The experimental results show that high-precision,quasi-real-time camera array calibration can be achieved with the dynamic calibration algorithm.