3D Reconstruction Of Large Object Based On Movable DLP Surface Grating Scanning

With the development of industrial modernization, the level of manufacturing has been greatly improved, it has become a trend that acquiring spatial 3D information in manufacturing. The 3D Measurement based on Structured Light has become popular for its character of non-contact, rapid response, wide measuring range, high precision, it is widely used in industrial production, cultural relics protection, reverse engineering, medical treatment, it's the most potential skill in the field of non-contact 3D measurement for its fast development and wide application.This paper aims to obtain 3D information in the process of robots stacking, designs two sets of 3D measurement system based on structured light to acquire the 3D information of objects. Complete forward and reverse gray code method and system calibration based on local homography are obtained to design the first 3D measurement system, complete forward and reverse gray code combined with eight-step phase-measuring method and system calibration based on radial fundamental matrix decomposition are obtained to design the second 3D measurement system. The research object focus on complex metal artifacts and flexible packaged rice. Some key techniques have been systematically studied, including system calibration, codification and decoding strategies, extracting valid regions, extracting the region of interest in point cloud, point cloud denoising, streamlining, registration, partial point cloud alignment, calculating the volume and surface area of the flexible packaged rice and so on.Firstly, this paper introduces the composition of the system based on structured light, derives the procedure of calculating the spatial 3D coordinate of the research objects, focuses on the study of system calibration, compares the effect of the commonly used system calibration with the calibration method based on local homography and radial fundamental matrix decomposition adopted in the paper by experiments. In terms of codification and decoding strategy, this paper emphasizes researching of complete forward and reverse gray code method and complete forward and reverse gray code combined with eight step phase-shifting method. The policy of extracting effective regions is based on optical component judgment. In the area of point cloud acquisition, using the established systems to acquire the point cloud of complex metal artifacts, comparing the effects. In the area of processing point cloud, a method is proposed to transform the 3D point cloud into the 2D image space through perspective projection and choose the region of interest, then deal with point cloud depends on response of data in 2D image space. In order to obtain the model of complex metal, this paper adopts multi-view registration of point cloud to register point cloud of complex mental from multiple different views and triangulate the whole point cloud data. This paper puts forward a method to make partial point cloud symmetrical to get the whole point cloud of flexible packaged rice. In the end, discussing the effect of various volume and surface area algorithms and testing various algorithms on flexible packaged rice point cloud, selecting the best one, designing experiments to verify its stability.The experimental results show that the systems designed in this paper with high precision, high stability, flexible operation, low cost, and meets the demanding of getting information of objects for robots stacking.