Research On 3D Mechanical Parts Reconstruction Based On Two Dimensional Image

In this paper,the 3D reconstruction based on binocular stereo vision as the research target.A binocular stereo vision platform is based on a mathematical model of binocular vision,and a 3D reconstruction system platform of two-dimensional images is designed.Then the key technologies are analyzed and studied deeply,its content mainly includes image preprocessing,edge detection,camera calibration,3D reconstruction process the specific contents are as follows:(1)Set up the hardware platform of binocular stereo vision,analyze the mathematical model of the binocular vision system in detail.The components of the system are analyzed;measures are taken to minimize the error and interference between the external environment and the system platform,so as to improve the accuracy and precision of each part of the 3D reconstruction.(2)A set of binocular stereo vision platform has been built,and taking full account of the light intensity and target object characteristics,completed the acquisition of the target image several times.(3)The relationship between the coordinate conversion of stereo vision camera,linear and nonlinear imaging model are analyzed.The classical edge detection such as Sobel operator,Roberts operator,Log operator,Laplacian operator and Canny operator are analyzed,and through the comparative analysis of the various operators,according to the research results and a new improved algorithm based on Laplacian operator is proposed.Through the experiments the effectiveness of the new algorithm in edge detection is verified.(4)According to the imaging principle of camera model,camera calibration method and the classification of commonly used classical calibration method are studied in detail: traditional calibration method,self-calibration method and Zhangzhengyou's plane template method.According to Zhangzhengyou's plane template calibration method,the internal parameters,external parameters and distortion coefficients are calculated.The validity of this method is demonstrated by experiment,and good accuracy of the camera internal and external parameters are obtained.These conclusions provide adequate protection for the follow-up smooth work and accurate 3D reconstruction.(5)Analyze the method of extracting feature points and basic principle of the Moravec algorithm and Harris algorithm.Through the experiments on the comparative study,it can be found that the Harris algorithm can effectively remove the less information and prone to aggregation of feature points,which shows that the Harris algorithm is effective in extracting feature points.At the same time,study on the matching of feature points extracted by the method of limit constraint,with high accuracy,short time.Through the use of binocular stereo vision system platform,the three-dimensional reconstruction process has achieved higher accuracy in each step,so as to achieve the object of three-dimensional reconstruction.Finally,through the study of two-dimensional images of three-dimensional mechanical parts reconstruction.using design platform of binocular stereo vision system,the obtained 3D reconstruction in the process of edge detection,camera calibration,stereo matching results and 3D model experimental reconstruction results,and the whole paper is summarized and Prospect analysis.