Automatic Reconstruction Of Three Dimensional Model Of Laser Machined Workpiece

As an efficient and practical cutting and welding method,three-dimensional laser machining has been widely used in automobile manufacturing,aviation parts processing and some other workpiece processing.Now manufacturing plants have been gradually trying to achieve full automation,which also means that any three-dimensional workpiece to be processed on the processing platform can be reconstructed as a 3D model to meet the accuracy requirements,which is of great significance for three-dimensional laser processing.Based on the above problems,a three-dimensional model reconstruction system of binocular structured light is built in this paper.Compared with monocular structured light technology,binocular structured light system only needs to calibrate two cameras once.In later use,in order to achieve better results,the position of the camera and projector can be flexibly adjusted according to the change of the scene without the need to calibrate again.The main work of this project is as follows:(1)By using complementary Gray code and maximum probability correction method,the boundary jump error in the process of unwrapping phase value by gray code combined with phase shift method is corrected,and by using fitting unwrapping phase value combined with table lookup method and double N-step phase shift method,the nonlinear effect of the projector caused by gamma transform is corrected.The two correction schemes reduce the corresponding errors and improve the reconstruction accuracy.(2)In this project,the efficiency of stereo matching is optimized by using two schemes of dividing interval,trial step size and parallel processing by CPU and GPU multi-threading,which reduces the total time of stereo matching from the order of minutes in serial processing to about 1s.(3)This project uses 4PCS algorithm for coarse registration of multi-perspective point clouds and ICP algorithm for refined registration of multi-perspective point clouds.And combined with PCL open source library to achieve visual point cloud registration and point cloud fusion module.(4)According to the multi-group schemes adopted in this topic,several groups of experiments were carried out respectively,and the measurement accuracy analysis and error analysis of the system were carried out.Under the working distance of 1.25 m,the plane fitting of the flat workpiece with the flatness of 0.01 mm was carried out.The maximum forward deviation distance was 0.467 mm.The maximum backward deviation distance was 0.375 mm.The average deviation distance is 0.016 mm,and the standard deviation is 0.127 mm.