Research On Detection Method Of Cable Surface Defects Based On Machine Vision

In recent years,as the demand for cables in my country has increased,the phenomenon of cable product quality substandard has gradually emerged.The form of manual visual inspection has gradually been unable to meet the growing production scale and accurate detection results.Machine vision technology has gradually been applied.In the field of product appearance quality inspection.The main research object of this paper is the surface defects of power cables with voltage levels of 10k V and above.Through the research on the light source,lighting method,industrial camera chip type and lens parameters in machine vision technology,the light-emitting diode(Light Emitting Diode,LED)strip light source is selected and the angle lighting method is used to complete the selection of the lighting part,and the area array gray scale industry is selected.The camera is equipped with a 25mm fixed focal length visible light lens to collect the original images of the three defects of cable dents,bumps and scratches.The background of the cable defect image is separated by the fixed coordinate method,and the relevant experiment of the cable image denoising is carried out.Experiments have found that mean filtering,Wiener filtering and median filtering have their own advantages and disadvantages for the processing results of two different types of noise.Aiming at the two kinds of noise can be produced at the same time,this paper proposes a new denoising algorithm and conducts mixed noise denoising experiments.Using the mean square error as a measure of the denoising effect,several filtering methods are compared,and the results show that the algorithm in this paper is better than the other three in denoising mixed noise.In the segmentation stage of the cable defect image,the edge detection segmentation is performed through three differential operators first,and the results show that the result of this segmentation method cannot meet the subsequent feature extraction criteria.By analyzing the gray-scale histogram of the defect,it is found that there is a significant gray-scale difference between the defect part and the background.Instead,three threshold segmentation techniques are used for experiments.The results show that iterative threshold segmentation can effectively segment the three types of defects.Then,the small interference items in the image are removed by morphological filtering,so that the cable defect image reaches the standard of defect feature extraction.The cable defect is traced based on the contour tracking method,and the defect area is marked by scanning the connected domain,and the area and perimeter characteristics of the defect are extracted,and the smallest size defect recognition experiment is carried out.The results show that the design in this paper can be more than 1mm~2Dimensional defects are identified.Based on the support vector machine,the classification training of the defect information is completed,and the result shows that the algorithm in the article has a comprehensive defect recognition accuracy of 93.3%.