Research On Detection Method Of Conveyor Belt Tearing In The Coal Mine Based On Machine Vision

As the core equipment of the coal mine transportation system,the mine conveyor belt is prone to tear damage failure after being crashed by sharp impurities such as bolts and schist in coal flow.To ensure the production safety of mining areas and improve the detection accuracy of coal mine conveyor belt tearing,the longitudinal tearing phenomenon is studied in this thesis.The conveyor belt motion blurred image restoration based on₀-norm regularization,the determination method of conveyor belt tearing based on curvature change analysis algorithm,and the determination method of conveyor belt tearing based on GLCM and multi-layer are proposed,and the conveyor belt tearing detection system is designed and implemented.The main work of this thesis is as follows:（1）Given the problem of motion blur degradation of conveyor belt images,Conveyor belt motion blurred image restoration based on₀-norm regularization is proposed.The motion blur image degradation model of mining conveyor belt is derived based on the general motion blur image degradation model.The₀-norm is used to depict the sparse characteristics of the blur image to estimate the point spread function.The suppression ringing effect model is introduced to weaken the image ringing effect.The comparison experiment verifies that the conveyor belt image restored by this algorithm has higher clarity,richer texture details and almost no ringing effect,the accuracy of conveyor belt tear detection model is improved by10.02%.（2）There are many interference factors such as dust,water stain,and light in the production environment of the mining area,so it is difficult to directly judge whether the conveyor belt tears through laser stripes.A determination method of conveyor belt tearing based on curvature change analysis algorithm is proposed.The linear characteristics of laser fringes replace the shape characteristics of longitudinal tearing,a preprocessing operation is designed to remove the redundant background information of the image to improve the image processing speed,and Steger algorithm is used to extract the center line of"one"laser stripe at the bottom of the conveyor belt.In the link of tear characteristics analysis,significant lacerations with the connected domain of the center line greater than or equal to 2 are first screened out,and then a more accurate curvature analysis is carried out on the center line with the connected domain of 1.According to the local jump and distortion of the center line curvature,the conveyor belt tearing fault is judged.Compared with the original model,the accuracy and detection speed of the proposed algorithm model are improved by 15.00%and 21.34%,respectively.（3）The small difference in the grain features of the"one"laser line at the bottom of mine conveyor belt in torn and non-torn states,it is difficult to train a deep convolutional neural network with high discriminative characterization ability,a determination method of conveyor belt tearing based on GLCM and multi-layer convolution feature fusion is proposed.This thesis represents conveyor belt image feature information from the GLCM feature and multi-layer deep convolution feature,giving full play to the complementary relationship between texture feature information and deep convolution feature information.The comparative experiments on CUMT-CID data set to prove that,this model has stronger feature representation ability and higher accuracy of conveyor belt tear detection.（4）The conveyor belt tearing detection system is developed.Relying on MVC architecture and adopting the hierarchical design idea,the data source,transport layer,infrastructure layer,platform layer,and application layer are designed respectively to realize real-time monitoring of conveyor belt health status.The software test proves that the system can meet the production demand of the mining area.