Research On Fault Detection Technology Of Conveyor Belt Based On Machine Vision

As the conveyor belt has been carrying materials for a long time and running with high load,the longitudinal tear failure of the conveyor belt often occurs.If it can not be detected in time,it will cause great economic losses,and even lead to safety accidents.In recent years,machine vision technology has been widely used in the research of conveyor belt longitudinal tear fault detection,but there are still problems of untimely and inaccurate detection due to a variety of complex factors in actual detection.Therefore,this thesis continues to study the conveyor belt longitudinal tear fault detection technology based on machine vision.The main research contents are as follows:(1)In order to solve the interference problem caused by the complex environment light on the surface of the conveyor belt when it is running,and the problem that the light intensity scattering of the laser stripe edge causes small noise on the edge,which affects the accuracy of centerline extraction,a longitudinal tear detection scheme based on the combination of line laser transmitter and filter is adopted,and an improved centerline extraction algorithm of laser stripe based on gray center of gravity method is proposed.First,the contrast of laser stripe is enhanced by using gamma correction algorithm;Then,Otsu method is used to segment and extract the laser stripe;Continue to use morphological operations to fine process the segmentation results;Finally,the center line of laser stripe is extracted by using the gray barycenter method improved by the dynamic programming algorithm,which realizes the accurate extraction of the center line of laser stripe and the determination of the position of laser stripe.(2)To solve the problem that the complex speed change,jitter change and surface curvature change during the operation of the conveyor belt affect the definition of the features in the generated conveyor belt surface feature image,a frame by frame splicing method of conveyor belt surface feature image based on SURF improved precision matching algorithm is proposed,and three splicing correction algorithms,speed correction,jitter correction and adaptive adjustment of laser stripe width,are proposed.Firstly,the feature points of adjacent frames of video on the conveyor belt surface are extracted;The feature points are matched with BBF algorithm based on K-D tree;Then,RANSAC algorithm is used to eliminate the mismatched feature points;Then the image translation transformation model is selected to complete the frame by frame splicing of the conveyor belt surface feature images;Finally,three correction algorithms are used to compensate for complex inter frame changes,and the longitudinal tear feature or other features in the conveyor belt surface feature image are clearly displayed.(3)In view of the fact that the complex surface feature factors of the conveyor belt during operation are easy to affect the accuracy of the identification of the longitudinal tear feature,a method for determining the longitudinal tear feature based on the template matching idea is proposed by taking the longitudinal tear feature as a new feature of the conveyor belt surface.Firstly,the binary image connected domain analysis algorithm is used to preprocess the features in the conveyor belt surface feature image for marking,denoising and repairing;Secondly,the SSDA algorithm,which improves the search matching strategy,is used to determine the new features of the conveyor belt surface;The slope and length characteristics of longitudinal tear feature are used to further determine the new features;Finally,it realized the timely and accurate detection of the longitudinal tear characteristics of the conveyor belt surface.