Research On Point Cloud Error Compensation Method For Stereo Vision Structured Light System

In recent years,stereo vision structured light 3D reconstruction system is widely used in measurement,reverse engineering,contour recognition and other scenes.It plays an important role in the manufacturing industry.The 3D reconstructed point cloud model is the data basis of post-processing and determines the accuracy of post-processing.Therefore,it is of great significance to improve the accuracy of point cloud reconstruction of stereo vision structured light 3D reconstruction system.However,there are several problems in the current research:(1)at present,the camera calibration methods are to establish the optimization objective equation,and then carry out iterative optimization to obtain optimal calibration parameters.Therefore,there are parameters estimation errors,which will lead to 3D reconstruction error.(2)In the process of imaging system structure calibration,the system parameters are not independent,but restrict and affect each other.The optimal system configuration also has errors,and the accuracy still needs to be improved.(3)Due to complex imaging steps and principle,it is difficult to establish a highprecision theoretical error model and error transfer model for stereo vision structured light 3D reconstruction system.To solve the above problems,a 3D point cloud error compensation method for stereo vision structured light 3D reconstruction system is proposed.The main contents are as follows:(1)The framework of error compensation system is designed.The whole error compensation system is divided into 3D error calibration module,3D error modeling module and 3D error compensation module,which defines the content and relationship of each module.(2)Aiming at the problem of comprehensive error calibration of stereo vision structured light system,a comprehensive calibration method for 3D error of stereo vision structured light system is proposed,which realizes the comprehensive calibration of the error considering imaging depth information in stereo vision structured light system.(3)Aiming at the spatial error compensation strategy,two methods of three-dimensional error model are proposed:inverse distance weighted interpolation method and neural network method.The error calculation of any point in the space is realized by establishing an error model.(4)The three-dimensional error compensation method for the stereo vision structured light system is studied.According to the imaging characteristics of the stereo vision structured light system,the established error model is applied to the point cloud error compensation method,which improves the 3D reconstruction accuracy of the stereo vision structured light system.(5)A software with 3D error calibration,3D error modeling and 3D error compensation functions is developed.The reconstruction accuracy of stereo vision structured light system is verified by using standard ceramic ball,and the ball diameter error and ball center distance error before and after compensation are measured.The experimental results show that after error compensation,the reconstruction error of stereo vision structured light system is significantly reduced,which verifies the effectiveness and feasibility of the framework and related algorithms in this thesis.