Research On Real-time Detection System For Cable Surface Defects Based On Machine Vision

As an important industrial wire product for transmitting electric(or magnetic)energy and information and realizing electromagnetic energy conversion,wire and cable are inevitably caused by factors such as production process,processing equipment and raw materials in the production process.Defects such as scratches,small holes,bulges,insulation damage,creases,and abnormal diameters directly affect the performance of the cable.The traditional manual detection efficiency is difficult to meet the detection requirements of high-speed production lines,and the missed detection rate and false detection rate are extremely high.Therefore,in order to meet the high speed and high precision requirements of product surface defect detection,the use of machine vision-based intelligent detection systems in the industry is increasingly necessary.This topic studies the surface defects of cables from the perspectives of hardware,algorithms and software of machine vision.A set of real-time detection system for cable surface defects based on machine vision is designed.The main research contents are as follows:Firstly,the visual hardware devices and software modules required for detection are designed according to the industrial production line environment and the characteristics of the cable.The image acquisition device can continuously capture high-definition images at a fixed frame rate of 360° at a fixed frame rate,and the maximum speed can reach 2 m/s;the upper computer interface of the software module can configure parameter information,display defect images in real time,and can Information such as the area,location and type of defects is saved for easy reference.Secondly,the cable image preprocessing algorithm is studied to improve the execution efficiency and processing precision of the latter algorithm,including median filtering noise reduction,homomorphic filtering enhanced image,diameter monitoring based on Canny edge detection,Hough linear transformation to extract cable area,The affine transformation corrects the tilt cable image and the like.Then,the suspected defect frame detection and defect segmentation algorithms are studied.The CV-Kmeans region classification method is introduced to establish an adaptive Gaussian filter background template,and the Pearson correlation coefficient between the original image and the template is calculated to quickly determine whether it is a defective frame.For the difference between the image containing the defect and the background template,the adaptive threshold segmentation algorithm and the open operation are used to extract the defects from the difference image,and the defects are searched,fitted,merged and labeled.Finally,the feature extraction and recognition of cable surface defects are studied.Among them,the geometric and texture features of the defect are extracted.The optimized BP neural network algorithm is used to identify and classify defects.Experiments show that the optimized BP algorithm improves the convergence speed and robustness,and the accuracy rate is 92.5%,which is better than the traditional BP neural network algorithm.In the system performance test,the experimental results also show good real-time performance.