Research On Conveyor Belt Surface Fault Location Technology Based On Machine Vision

As an important transportation equipment of modern production in China,belt conveyor is widely used in coal,metallurgy,mining,ports,chemical industry and many other industries.The conveyor belt is an important part of the conveyor.It is prone to scratches,longitudinal tear,damage and other surface failures,which is seriously endangering the safety of production.At present,the machine vision technology has been applied to fault detection of conveyor belt.However,the repair process to locate the fault will spend a lot of time,manpower and material resources due to the head-tail connecting and cycling operation,even if faults are detected by use of machine vision technology.Therefore,faults location becomes an important technical problem for belt visual monitoring.A fault location method based on machine vision is proposed to solve the problem of surface fault location for belt visual monitoring.According to the characteristic that numbers are easily identified by maintenance personnel,numbers are branded on the edge of upper and lower surface of conveyor belt,belt images are processed to achieve position and recognition of the marker number,and to locate the conveyor belt surface faults.For location of digital markers,a new algorithm based on visual saliency and convolutional neural network is proposed.Firstly,the collected images of the belt marking are processed by the gray linear transformation to enhance the contrast of the images.Then,a spectral residual approach is used to obtain visual saliency map.Secondly,an improved histogram threshold segmentation algorithm is designed to segment the belt images with tag saliency map,which is used to position digital signatures preliminary.Finally,the convolution neural network is used to classify the original image and to distinguish between marked numeric and non digital regions.The experimental results show that the proposed method can be well positioned for the labeled figures on conveyor belt images.For the identification of a digital signature,a number recognition algorithm of belt conveyor based on deep learning is proposed.The AlexNet model is improved to design an appropriate network structure parameters to deal with belt mark images.Experimental results show that the algorithm has high recognition accuracy.On this basis,the belt surface fault location modular based on machine vision is developed.The positioning modular is tested by the visual on-line monitoring system developed in the laboratory.The test results show that the module is reliable.