Research On Moving Coal Particle Detection And Tracking Based On Machine Vision

As the pillar energy in China,coal is an important raw material for industrial production,which is closely related to the construction and development of the country.Improving the comprehensive utilization rate of coal meets the current development needs of the country.Through the screening and classification of coal,we can get a variety of coal products,and expand the use of it.The vibrating screen for coal is the key equipment in the coal surface separation process.Its screening efficiency directly affects the utilization rate of coal,which is of great significance to reduce resource consumption.In view of the research on the movement characteristics of coal particles in the screening process,it can guide the optimization of the structure design of the vibrating screen and the setting of corresponding parameters,so as to improve the screening efficiency.The complex movement of coal particles on the screen can be well reflected by the actual measurement experiment.However,the appearance of coal particles is randomness,and the movement speed on the screen is fast.Besides,it's easy to block,break and depolymerize during the screening process,which brings challenges to the actual measurement experiment.Therefore,this paper proposes a method based on machine vision to detect and track the moving coal particles,which provides a new idea and method for the experimental study of the coal particle motion characteristics on the vibrating screen.First of all,a simulation screening vibration experiment platform is built in this paper.In view of the difficulty of the ordinary camera to collect the video images of the coal particles moving at high speed,the non-contact high-speed camera is selected as the collection method of the coal particles moving image.This paper analyzes the kinds of noise that are easily occurred in the video sequences of coal particles,and studies several commonly used noise removal preprocessing methods.In this paper,the guided filter with edge preserving is selected as the noise removal method,and the morphological processing method applied to foreground mask is analyzed.Secondly,three frame difference method of adaptive threshold,Vibe algorithm,SGM and EM-GMM are used to detect the moving coal particles.By comparing the experimental results,EM-GMM can better suppress the shadow generated during the movement of coal particles and adapt to a certain degree of disturbance.Compared with other methods of moving object detection,the foreground area of coal particles is more complete.Finally,the follow-up tracking is carried out on the basis of moving coal particle detection.Because the coal particles are easy to be occluded in the process of moving,the Mean-Shift tracking algorithm based on template matching will lose tracking when the moving coal particles are occluded,while the Kalman filter tracking algorithm based on the motion state estimation is characterized by the centroid position of the foreground area detected.Through the state prediction and update,it is more robust for tracking in the case of occlusion.In order to track multiple coal particles,this paper proposes a tracking mechanism based on the fusion of Hungarian algorithm and feature Kalman filter.The Hungarian algorithm is used to solve the minimum loss matrix of the centroid position of coal particle detection and tracking prediction,and the assignment and tracking of multiple coal particles are realized.And it can track the wet coal agglomerates stably when they are broken,and different coal particles are blocked after depolymerization.