Three-dimensional Reconstruction Of Moving Target Based On Multi-sensor

With the development of computer stereoscopic vision,the application frequency of 3D reconstruction technology in video surveillance,video animation,virtual reality and other fields has become more and more frequent.The multi-sensor 3D reconstruction technology is a highlight of scientific research.Although many problems are faced in the process of 3D reconstruction,its widespread applications prompt many scholars to persevere in related research.In this paper,we focus on the camera calibration technology,stereo matching algorithm and 3D reconstruction algorithm based on the three-dimensional reconstruction of multisensor moving targets.The 3D reconstruction images are displayed by OpenGL.Finally,a host computer system based on MFC is established to realize convenient human-computer interaction.The main research contents and innovations are as follows:In the aspect of camera calibration,the coordinate systems involved in 3D reconstruction system and their transformation relationship are introduced firstly.The geometric model of multi-sensing vision system is introduced.Based on the analysis of the calibration principle and process of Zhang Zhengyou calibration method,Large field of view for the separation of the calibration algorithm are introduced.Small checkerboard composed of large field of view calibration board,the center of the small checkerboard coordinate system is set as the absolute coordinate system,the use of Zhang calibration method to obtain the internal and external parameters.Finally,the conversion matrix is calculated by Rodrigues parameter method,the coordinate system is unified,and the external parameters of the camera are obtained.The error of calibration accuracy is about 0.2 pixel.In the aspect of stereo matching,we focus on the steps of stereo matching algorithm,and abstract the objective function of cost aggregation as the weighted least squares problem,and add the Tikhonov regularization term to it for consistency constraint.The objective function is optimized by using Gaussian pyramid form.The proposed stereo matching algorithm uses the local stereo matching algorithm to obtain the multi-scale cost entity through the cost aggregation process.Combined with the greedy rule of the local stereo matching algorithm,the disparity map is obtained by optimizing the global search domain.In the moving target three-dimensional reconstruction,we focus on the point cloud 3D reconstruction and triangulation.Firstly,3D coordinates of the space point cloud are calculated according to the 3D parameters obtained from the camera calibration and the stereo matching.Then the Delaunay triangulation is performed on the 2D image and the 3D reconstruction image is displayed by using OpenGL.In order to validate the effectiveness of the algorithm proposed,we design a host computer software system based on MFC,including five functional modules,and test and analyze each module to achieve convenient people Machine interaction.