Research On Key Technologies Of Coal And Gangue Intelligent Sorting System Based On Near-infrared And Depth Information

Coal is one of the most important sources in China.The separation of coal and gangue is a key link in coal mining,and its intellectualization promotes the intelligent development of coal mining.However,it is difficult to sort coal and gangue quickly and accurately because of some reasons,such as fast running speed of belt,different shapes and sizes of gangue,large quantities of coal and gangue,and so on.To solve the problem,an intelligent sorting system for coal and gangue is proposed based on multi-information fusion.The system uses the performance advantages of near-infrared camera and deep learning to identify coal and gangue,and uses the fusion of near-infrared and depth camera to obtain the type,depth and pose information of gangue.Based on the information obtained by the camera and the flexible sorting model built in this paper,the function of rapid and accurate sorting of different gangues on the high-speed belt is realized.The specific research work is as follows:(1)An experimental platform for intelligent sorting of coal and gangue is designed and built.The platform consists of preprocessing module,coal gangue identification and pose acquisition module and flexible sorting module.The pretreatment module is used to screen and separate the initial large quantities of coal and gangue,so as to prevent the overlap of coal and gangue,which is convenient for the subsequent visual module to obtain the information of coal and gangue.By analyzing the performance of different sensors,the module of coal and gangue identification and pose acquisition adopts nearinfrared camera instead of ordinary camera to collect and identify the image of coal and gangue,and uses the depth camera to obtain the pose information of gangue.Using the information obtained by the camera,the flexible sorting module adopts the self-designed array pneumatic separation device combined with the flexible sorting model to realize the flexible sorting of coal and gangue.(2)In view of the traditional machine vision algorithm can not solve the problems of image noise,complex background,difficult feature extraction in dynamic scene,which is difficult to meet the complex environment and real-time requirements of coal mine site,this paper proposes an improved YOLOv3 algorithm for coal and gangue identification.Based on the YOLOv3 network,the algorithm improves it in three parts,which shortens the prediction time of network model,improves the recognition accuracy of coal and gangue in dynamic scene,improves the matching degree between prediction box and target object,and makes it more suitable for gangue sorting in complex scenes.After obtaining the type and coordinate information of coal and gangue by near infrared camera,it is necessary to match the depth information obtained by depth camera.Therefore,this paper obtains the conversion matrix by camera calibration.Through matrix conversion and image alignment,the information of coal and gangue obtained under two cameras can be effectively matched.(3)The size and posture of gangue vary greatly,so it is difficult to sort each gangue to a suitable position with single sorting strategy.Besides,this paper analyzes the factors that affect the trajectory of gangue before and after pneumatic separation,deduces the theoretical formula of horizontal movement distance of gangue pneumatic separation,establishes the mathematical model of gangue flexible sorting,and carry out experiments of pneumatic separation with different shapes and sizes of gangue.Experimental data is used to realize error compensation of the model.The modified sorting model has important guiding significance for pneumatic separation of coal and gangue.(4)Finally,through the experimental platform,combined with the above coal gangue identification algorithm and coal gangue flexible sorting method,the intelligent coal gangue sorting system in this paper is tested and verified.Comparative experiments show that the coal gangue identification algorithm proposed in this paper can effectively improve the detection speed and identification accuracy of coal and gangue.In dynamic scenarios,the identification accuracy of coal and gangue can reach 92.5%,and the average detection speed only needs 37ms.The experiment also shows that the flexible sorting model of gangue is effective for improving the sorting accuracy of gangue.Under the same experimental conditions,the sorting accuracy can reach 94.66%by using the combination of flexible sorting model of gangue and array pneumatic separation device.The intelligent sorting system of coal and gangue can accurately identify and sort coal and gangue of different shapes and sizes under high-speed belt movement.