High Precision Measurement And Defect Detection Based On Point Cloud

Since introduced in the mid-1990s,3D scanning technology has getting more and more increasing attention from industrial production research neighborhoods.The three-dimensional scanning technology can obtain the three-dimensional point cloud data of the surface of the measured object in the camera coordinate system through high-speed scanning measurement.It has the characteristics of high speed,high accuracy,digitization,real-time dynamic display,and non-contact.However,a single scan of the 3D scanning technology can only obtain point cloud data of the object from a single side.So,the point cloud data obtained from multiple measurements needs to be processed to obtain the complete 3D data of the detecting object.And because point cloud data is disordered and sparse,processing point cloud data is quite difficult.In the neighborhood of industrial production,real-time monitoring and testing of product production is one of the indispensable links.Conventional product inspection usually uses machine vision inspection technology.It is difficult to complete high-precision measurement of industrial parts based on image processing for part defect detection,so it is impossible to detect parts such as surface deformation and dimensional errors.Therefore,non-contact high-precision defect detection of industrial parts is a recognized problem in the industry.The non-contact and high-precision defect detection of industrial parts mainly includes:part point cloud acquisition,complete 3D data acquisition of detecting part,part category pose determination,and part defect detection.The main work is as follows:First,in order to solve the problem of part point cloud data acquisition,a three-dimensional scanning system based on structured light coding bas been built.The system was used to obtain high-precision point clouds of the detecting parts,providing a data basis for subsequent point cloud processing.Secondly,since a single 3D scan can only obtain single-sided point cloud data of the detecting part,in order to obtain the complete 3D data of the detecting part,a method of rough and fine stitching of point clouds based on 2D-3D neighborhood information is proposed.The two-dimensional image corresponding to the cloud is used to stitch the three-dimensional point cloud data,and the point cloud stitching after collecting the detecting parts from multiple perspectives is obtained to obtain the complete three-dimensional data of the detecting parts.This provides the basis for extracting the semantic and pose information of the parts to be detected from the point cloud data.Then,in order to judge the pose of the parts to be detected,a point cloud coding method is proposed.The disordered point cloud is encoded into fixed-length and ordered structural data,and the disorder of the point cloud is solved to a certain extent.Obtaining part category and pose information from cloud data provides a basis for part defect detection.Finally,in order to verify the effectiveness of the three-dimensional part processing method proposed in this paper,a defect detection standard for parts to be tested is proposed,a point cloud-based three-dimensional point cloud defect detection system is developed,and a defect detection experiment is performed on the assembly of the part to be inspected.Complete the high-precision measurement and defect detection based on point cloud.