Design Of Belt Deviation And Sprinkling Detection System Based On Machine Vision

Belt conveyors are widely used in metallurgy,mining and other industrial production industries in the transportation links.If the belt conveyor works for a long time,it is inevitable that there will be problems of misalignment and spillage,if it is not solved for a period of time,it is easy to cause more serious failures,affecting production and accompanying safety hazards.Manual inspections are inefficient and problematic.However,at present,belt fault detection pays more attention to the problems of belt tearing,belt breaking and mistracking,and does not pay too much attention to the situation of spilling,ignoring the equipment damage and safety problems that may be brought by belt spilling,and sprinkling is related to belt mistracking.Therefore,this thesis combines machine vision to study belt misalignment and spillage,and designs a detection system to detect belt misalignment and spillage with high precision.The main work of this thesis is as follows:(1)Due to the limited clarity of the image acquired through the lens,it is necessary to highlight the characteristics of the belt and the material it carries.In this thesis,a set of retinex’s Multi-Scale Retinex algorithm is designed to first do image enhancement,and then combine image grayscale and image filtering methods to process images,which can highlight important features related to belts.(2)In order to solve the problem of belt edge positioning,a method combining the difference between belt and background features with the Canny operator-Hough transform is proposed to accurately locate the edge positions on both sides of the belt.In order to solve the problem that it takes too long to detect belt mistracking based on deep learning,a mistracking detection model is established by using the positioned belt edge position information,and the corresponding early warning of the belt mistracking degree is realized by setting different thresholds.The belt mistracking detection method proposed in this thesis can give a graded warning of the degree of belt mistracking,which takes less time and is more practical.(3)Aiming at the problem of low detection accuracy and small adaptation range of traditional threshold method,combined with the edge position information on both sides of the belt to obtain the change of gray value of each pixel on the section line perpendicular to the running direction,this thesis proposes an SVM(Support Vector Machine)spilling detection model combined with the belt and the material distribution characteristics it carries to detect whether there is a spilling risk on the belt.Experimental data prove that the spilling detection method has strong generalization ability,which can well prevent a series of problems caused by material spillage,and can prevent the occurrence of belt misalignment from the perspective of material distribution.(4)Combined with actual needs,use the LabVIEW platform to design the functional module of the proposed belt misalignment and spillage detection,build a software system,and output and display the real-time detection results,and assist the staff to judge the working status of the belt in combination with real-time images.In this thesis,the system is tested using the video image of the belt conveyor used in enterprise transportation,and the results of multiple experiments show that the overall detection accuracy of mistracking and spilling is above 98%.Therefore,the belt misalignment and spilling detection system proposed in this thesis takes a short time and has high precision,which can achieve the purpose of saving costs and ensuring safe production.