Research On 3D Recognition Of Component Based On Digital Grating

Manufacturing industry is very significant in China.According to statistics in 2018,manufacturing GDP accounts for 29.4% of the total GDP.At the same time,manufacturing industry is indispensable in the research and development of new technologies,economic development and even international competition.Digitalization is the bottleneck restriction of manufacturing automation.Three-dimensional reconstruction technology is widely applied in reverse engineering,quality inspection and other fields,and it is the basis of workpiece digitization.The traditional workpiece measurement method is limited by the difficulty in achieving a balance between accuracy,efficiency,and cost,as a result it is difficult to achieve widespread popularity in the industry.This paper constructs a structured light vision inspection system from two aspects of hardware and software,studies the equipment selection and system calibration,the generation of point clouds from object surface and the subsequent processing of point clouds,such as segmentation,registration and visualization,and A series of experiments are conducted to verify each phase of the system.This article first selects hardware devices based on the application scenarios of the system.The camera imaging principle and the projector-camera geometric relationship are modeled to obtain the phase-height mapping.The process to calibrate the camera intrinsic parameters and mapping relationship parameters was designed.Experiments indicated that the method is simple and fast,and the relative position between camera and projector is not specialized.Secondly,the reconstruction of the object surface 3-D topography is studied.A four-step phase shift method is applied to calculate the phase of the modulated fringe on the object surface.An improved quality guidance method and the multi-frequency heterodyne method are applied to expand the phase.Based on the pinhole principle,the relationship between the phase and the 3-D shape of the object is derived.The experiment proves that the above algorithm can complete the 3D modeling of complex surfaces.Meantime it is concluded that the former method is more effective in a small range,and the latter method is better in a large range.Then The 3-D point cloud post-processing is studied.Since the target point cloud is mixed up with the background,the OTSU algorithm is combined with the Kmeans algorithm to segment the target point cloud from the background.,which effectively solves the problem of local convergence.Considering that large-scale objects cannot be measured in a single operation,the FPFH descriptor is applyed to extract point cloud features,the RANSAC algorithm performs coarse point cloud matching,and the ICP algorithm performs precise matching for point cloud registration.Combined with application scenario,experiments show that the algorithm can be addopted for point cloud generated from complex scene.Aiming at the problem of poor visualization of point clouds,the point cloud surface is reconstructed using Poisson reconstruction algorithm.Experiments show that this method can effectively characterize the local characteristics of point clouds.In the end,C++ language is adopted to develop the software system on Visual Studio 2017.The system calibration,3-D reconstruction,point cloud post-processing and other functions are encapsulated in different models.At last,a simple GUI is established.