The Design Of The Visual Detection System For Wide-format And Continuous Surface Defects

This article describes a visual inspection system of wide,continuous surface defects,which uses a plurality of image processing subsystem collaborative to complete the inspection task.For continuous operation characteristics of the measured object,proposed the detection scheme based on linear array camera;for wide surface characteristics of the measured object,proposed the multi-camera stitching scheme to complete the inspection task.For the wide continuous surface defect detection system,we give the data processing program.In order to quickly extract the defect target,in the stage of the image acquisition,according to the optical reflection characteristics of the tested surfaces and defects,execute the illumination mode and light structural design to highlight the defects target under wide scene,using line scan cameras with linear light source detection system to collect the initial image.Combing the detection mode of intelligent camera which is based on FPGA and DSP and the processing mode based on server and client,this paper proposed a multi-core processing method on the distributed visual inspection system.Based on the client and server mode,describes the detection scheme of the C / S work system,and gives the workflow in the C / S system,each piece of camera configures image acquisition and processing card based on FPGA + DSP architecture,the processed defect information send into CMP computer to further image understanding and analysis,every two information acquisition and processing card configure one CMP computer,each CMP computer send the finally image data and detection information into the server to store and display;for the data communication problem about the multiple detection subsystem and server in the detection system,this paper designed a multi-point to point image transmission system to realize multiple detection subsystem transfer defect data to the server.According to the characteristics of the image data processing and applications,proposed the arrangements of serial-parallel processing structure about data and algorithm,and from the angle of dimension reduction and algorithm structure analyzed that how to improve the speed of data processing;on the other hand proposed the timesharing and multi-core processing scheme of image date in the visual detection system,to realize the image data processing of wide surface efficiently,improve the speed of the detection system.For wide continuous surface,the test area is large,this article uses multi-line camera stitching in the width direction to constitute a wide field of view.For the format motion orientation(vertical)segmentation it is proposed a format split scheme based on the location of the defect;for the width direction(horizontal),the split decided by the camera's position when the camera stitching,and analyzed the size of the overlapping area of the field of view about adjacent camera as well as the issue of the defects determination in the overlapping area.According to the given data processing structure,this paper studied the hierarchical image processing algorithms for multi-processor.It realized the median filtering and sobel edge detection in the FPGA,used the segmentation based on edge realize the defect segmentation in DSP,completed defect feature extraction and defect classification in the CMP computer.The experiment verified the feasibility of the hierarchical image processing scheme for multi-processor.