The System Design Of Appearance Defect Detection System For The Melting Point Of The Front Panel Of The Crate

Machine vision technology is concerned with image processing,feature localization,pattern recognition and other technologies,and is used in industrial production,inspection and other fields.With the popularity of the Internet,computers have become an absolutely essential part of people's lives,and then the quality inspection work for computer host panels is also of special importance.The object of this study is the detection of melting point defects in the front panel of the chassis.The machine running cycle on the assembly line is long,the hot melter will inevitably leak,it is difficult to ensure the good product rate of the workpiece.To solve this problem,most factories use manual testing.For non-stop pipeline operations,workers inevitably fatigue,it leads to misjudgment,and the high labor cost and low effectiveness make it difficult to meet the high volume of work.Therefore,the introduction of machine vision into the assembly line,after the hot melt to increase the visual inspection station,not only can meet the advantages of high efficiency,high quality of the assembly line,but also effectively liberate productivity,speed up the process of plant intelligence,effectively solve the above-mentioned problems.This paper presents a more detailed study of the front panel melting point defect detection algorithm for chassis and improves the application of the related algorithm.Due to the low contrast of the pictures taken by the industrial camera,it is difficult to distinguish the areas of workpiece and background,and more difficult to extract the melting point area in the workpiece area.Therefore,In this manuscript,We put forward a low-contrast image enhancement method based on wavelet variation and second-order discrimination,which uses the second-order discrimination in wavelet reconstruction to control the proportion of edge components in the high-frequency part of the image and improve the image contrast while preserving detail information.The superiority of this enhancement method is verified by histogram balanced image enhancement.Laplace image enhancement,image enhancement based on homological filtering,and image enhancement based on contrast constraint adaptive equalization method.The image enhancement results are split,the grayscale threshold of the melting point area is set,and the melting point can be initially determined by measuring the ratio of the partitioned gray histogram to the threshold value by comparing to determine the possible area and perimeter of the melting point,and to further determine the melting point area.By comparing the characteristics of defect melting point and non-defect melting point,based on BP neural network training sample,input the contour number,edge density,grayscale characteristic value of melting point area,area of defect area,perimeter and other 5characteristics,the melting point is fully melted,semi-melted or unmelted,and the output results are output.The project relies on the integrated environment of Visual Studio 2017 to develop a melting point defect detection system.Through network communication with PLC,the whole automation system is connected.And through the industrial field practice,it is shown that this system can meet the requirements of high-volume operation efficiency,real-time,and the correct rate to meet the requirements of the factory.