Surface Defect Detection Based On Line-Structured Laser Scanning

With the rapid development of manufacturing industry,product quality inspection standards are more and more strict,more and more enterprises begin to pay attention to the detection of product surface defects,and for the detection of product surface defects,most manufacturers rely on manual detection,the method of manual detection is low efficiency with high error rate and high cost.so the efficient detection method based on computer vision is more and more valued.This method obtains the two-dimensional image of the product surface and analyzes and detects the two-dimensional image information.However,in many cases,because the depth information of the target cannot be obtained from the two-dimensional image,some defects may be difficult to be detected,and the three-dimensional detection method can obtain the depth information of the target,reconstruct the defects that could not be identified previously.Three-dimensional detection is more reliable and intuitive,so three-dimensional surface defect detection obtains more attention.In this paper,an experimental design of the line-structured laser scanning system is carried out.Based on this system design a calibration process.The point cloud data obtained by the system is combined with an appropriate method for defect detection and quantification to achieve the purpose of defect detection.In this paper,the following work is carried out in surface defect detection based on linestructured laser scanning:Firstly,The detection in this article is mainly to detect deformation information,so building and verifying an effective acquisition system is the first task.Study the principle of visual system based on line-structured laser.The principle is to increase the conditional constraint through the line-structured laser to assist the camera to complete the depth information acquisition.The general model of the line-structured laser vision system is studied.Because the general model is analyzed in the world coordinate system,it is difficult to accurately obtain the relative pose of the camera in this reference system.Therefore,the reference coordinate system is established in the camera coordinate system for analysis,and a simplified model is obtained.On this basis,the corresponding algorithms and steps for the calibration of the light plane and the speed direction of the conveyor belt are formulated.At the same time,some methods and indicators are proposed to measure the accuracy of the calibration.Through repeated experiments and adjustments to the experimental platform and algorithm,an effective calibration method and an effective vision sensor were finally obtained.Secondly,this paper adopts the method of mapping the point cloud data to a two-dimensional height grayscale image.By using appropriate interpolation methods,we can obtain a complete two-dimensional heightmapped image and then combines appropriate image processing and segmentation algorithms for defect detection.After detecting any defects,according to the characteristics of the system scan,the point cloud data acquired is subjected to frame-by-frame defect detection.Because the single-frame detection cannot determine the relationship of the frames.Based on this,it is proposed to merge and cluster the defect points detected in a single frame by using a parallel search method.This method is different from common clustering methods,and is not affected by the initialization setting and does not need to know in advance how many defect sets.This simplifies the clustering difficulty.Combined with certain dimensionality reduction processing,the defect point set was effectively extracted.Finally,triangulation is used to quantify the defect set.Thridly,A three-dimensional scanning defect detection experimental platform was established,and the detection software was developed under the Visual Studio 2017 software development platform to realize the real-time scanning visualization function,while displaying the scanned local point cloud and color height map image.The defect detection of metal profiles was carried out by using this experimental platform.The experimental results show that the method has strong feasibility.