3D Visiual Measurement Algorithm For Complex Industrial Parts

Complex surface with irregular curvature a kind of surface with unstructured shape and abrupt curvature,and its edge is difficult to detect.As an important part of aeroengine,aeroengine turbine blade is a typical complex industrial part.Therefore,3D point cloud processing of aeroengine blade is an important step of 3D measurement,which has an important impact on the accuracy of subsequent machining and grinding.In this paper,the algorithms of point cloud preprocessing,segmentation,smoothing and reconstruction are studied,mainly to solve the problems of low accuracy,poor real-time performance and robustness in the measurement and machining process.The contents of this paper are mainly composed of the following parts:1)Aiming at the measurement demand of high-end intelligent manufacturing automation production line,3D measurement platform for complex industrial parts is built,and software module and processing flow are designed.Among them,the 3D measurement system includes high precision positioning subsystem,measurement subsystem and multi-robot cooperation subsystem.The software module is mainly divided into software layer,function layer and application layer.The precision machining flow is divided into the process of workpiece feeding,point cloud acquisition,down sampling and filtering,point cloud segmentation,smoothing and reconstruction,grinding and measurement steps.2)The point cloud preprocessing of complex industrial parts is researched.Due to the inherent defects of the equipment,background and other moving objects,the obtained blade point cloud has large numbers of outliers and noise points.In this paper,a voxel growth based filtering algorithm is proposed,which filter s the background points effectively and ensures the accuracy of subsequent segmentation and reconstruction.At the same time,the amount of blade point cloud is greatly reduced and the efficiency of 3D visual measurement algorithm is improved by voxel rasterization.3)A novel point cloud segmentation method based on singular value decomposition and sub-cluster merging is proposed,which solves the problem of difficult segmentation and false segmentation of complex industrial parts.By establishing pairwise connections based on the local flatness and normal deviation minimum,the seeds of the sub-clusters are determined,and the sub-clusters are merged via the curvature deviation and adjacent criterion of bounding box.After experimental comparison of the proposed method and RANSAC,boundary extraction and region growth algorithm between the actual and simulated blades,the proposed method filter noise and outliers and retain the effective points of the blade,which greatly improves the segmentation accuracy and efficien cy of the blade point cloud.It can be effectively applied to precision measurement and machining of complex industrial parts,which provides strong data support for subsequent measurement,machining andrinding steps.4)Aiming at the problem of the holes and unsmooth,the moving least square(MLS)method is used to fill the holes and smooth the point cloud.And an improved fast delaunay triangulation algorithm is proposed,which builds triangular mesh based on edge forward growth and establishes data set partition,avoiding the problem of low efficiency caused by conventional voronoi graph reconstruction.The effectiveness and rapidity of the smoothing and reconstruction algorithm are verified by comparing the before and after smoothing with the simulated and actual blade reconstruction effect and the cloud operation time of different input points.