Research On 3D Reconstruction Algorithm Based On Binocular Vision

With the rapid development of science and technology,Binocular vision,which is an important branch of computer vision,has been greatly improved and gradually applied in various fields of production and life.Like the human visual system,the camera captures the image pairs of the space object from different angles to calculate the disparity of the coordinate points of image pairs,which correspond to the same spatial point.Finally,the3 D information of the real scene is obtained from the 2D image.Binocular vision includes image acquisition,camera calibration,image preprocessing and feature points detection,stereo matching,3D reconstruction.This article focuses on the camera calibration,stereo matching,and 3D reconstruction of binocular vision.The main contents are as follows:First of all,it systematically outlines the research background and significance of the topic,and describes the research status of camera calibration,stereo matching and 3D reconstruction described.It also introduces the key issues of the three-phase study.All of these provide the theoretical basis of the follow-up research.Secondly,in the camera calibration section,traditional binocular camera calibration method is replaced by the neural network,which has the ability to strongly approximate nonlinear mappings.Mainly aiming at the disadvantage of neural network training falling into local minimum value,the hybrid algorithm h-HS-HuS is used to optimize the neural network to realize the global optimization of space.Simulation experiments have verified the effectiveness of the improved algorithm.Thirdly,the image preprocessing and feature point detection are performed,before the image stereo matching.Filtering the noise in the image,histogram equalization brightness difference and Wallis sharpening image preprocessing,the Harris algorithm detects corners.At the stereo matching stage,the dynamic programming stereo matching algorithm based on multiple similarity measure fusion is proposed to calculate disparity,which combines the local stereo matching and the global stereo matching algorithm.The merging algorithm ensures the matching accuracy and the real-time performance,which lays the foundation for the 3D reconstruction.Finally,in the 3D reconstruction part,the scattered space points are first calculated,and then an architecture,which is obtained through subdividing the spatial points by the improved Delaunay triangulation method,is texture mapped to finish 3D reconstruction.The overall process of 3D reconstruction based on binocular vision is completed through experiment simulation at last.