Parking Scene Reconstruction Based On Binocular Vision

With the rapid development of virtual reality and 3D modeling technology in China,computer vision technology has been rapidly developed and gradually applied to people's daily life.Moreover,3D reconstruction technology is a very popular field in computer vision.In the face of the parking scene,how to efficiently and accurately perform 3D reconstruction become a key issue.In front of the scene when parking,how to efficiently accurate 3D reconstruction become a key issue.For the reconstruction task of parking scene,the spatial information is mainly collected from different angles,and then the parallax of coordinates in the image pair is calculated,so as to obtain the 3d information of parking scene.Using binocular vision for 3D reconstruction is generally divided into camera calibration,image prepossessing,feature point extraction,stereo matching and3 D reconstruction.By studying the principle of binocular vision and deeply exploring and analyzing the parking scene in the background of parking,the parking scene reconstruction method based on binocular vision is constructed by flexibly applying the 3d reconstruction method to parking scene.The binocular vision and its three-dimensional reconstruction from the following three aspects.First of all,the traditional scale-invariant feature transform(SIFT)in image matching contained a large number of calculations and dimensions which led to the problem of large amount of computation and long matching time.Therefore,an improved algorithm based on PCA-SIFT was proposed.The algorithm used a circular descriptor to reduce the dimension of SIFT.At the same time,the principal component analysis method was used to reduce the dimension of the descriptor,so as to eliminate the dimension of the descriptor.Secondly the Euclidean distance and cosine similarity functions were optimized by the stratified particle swarm optimization algorithm,and the eigenvalues were obtained according to the difference,so as to find the extreme value of the function and obtain the matching point.Finally,by collecting binocular images of cars,SIFT algorithm and PCA-SIFT algorithm were used to carry out feature matching for the collected images,and sparse point cloud was established for 3d reconstruction of parking scene.Experimental verification and results analysis of two reconstructed images were carried out to verify the effectiveness of the proposed algorithm.Figure 26;Table 6;Reference 55...