Design Of 3D Reconstruction Software Based On Binocular Vision

Computer technology has been active in all aspects of life.Computer vision has become a hot-spot nowadays.And the technology of 3D reconstruction in computer vision is an important research technology in computer vision.Meanwhile,binocular vision is a very important technique for 3D reconstruction of the scene.It uses two or more two-dimensional images taken by a camera to calculate the three-dimensional information.The design of 3D reconstruction software based on binocular vision is to achieve three-dimensional reconstruction,from image acquisition to stereo matching,depth calculation,point cloud processing,and final triangulation.Besides,it is to provide an extensible three-dimensional reconstruction platform,which is beneficial to a variety of algorithms to achieve three-dimensional reconstruction.In the design process,combined with the design mode,we can choose the algorithm of each process according to the need.Based on binocular vision,three-dimensional information reconstruction software is designed,using strategy pattern to improve the extensibility of the platform.Each function module provides an interface for different algorithms to be implemented through the interface,and the algorithm classes are independent and independent of each other,and each interface class provides the required basic data and methods,reducing the amount of software code.The 3D information reconstruction software based on binocular vision mainly includes five functional modules: image acquisition,stereo matching,depth determination,point cloud processing and triangulation.Two cameras are used to get scene images from different perspectives,and the 3D reconstruction software reads the two-dimensional scene image.The left and right two matched images are stereo matched,and the Disparity graph of the stereo matching result is optimized and denoise.The disparity information is calculated with the principle of triangulation,and the camera model is used to calculate the image.The 3D point cloud information of the scene is obtained,when calculating the point cloud coordinates,the method is used to calculate the three-dimensional scene through the disparity image pixels,and there is a lot of redundancy in the point cloud.Three dimensional reconstruction software uses Canny edge detection method to constrain the point cloud data and keep the scene contour.And a shortest distance triangulation algorithm is designed.Finally,the experimental results and existing problems are analyzed.From the experimental results,3D reconstruction software can basically realize the process from stereo matching to 3D reconstruction.