Research On Method Of Identifying Foreign Object On Belt Conveyor Based On Deep Learning

As a large coal producer and a high coal consumption country,the research on the transportation process of coal is very important.At present,coal is mainly transported through belt conveyors in industrial production,foreign object mixed in coal will not only reduce the quality of coal,but may also damage the belt conveyor.Therefore,when foreign object enters the belt conveyor,it is especially necessary to accurately identify it in the early stage.Due to the continuous movement of the belt conveyor,it is difficult to meet the real-time requirements for the foreign object identification method on the belt conveyor.To solve this problem,the YOLOv3 model is improved and applied to the detection of non-coal foreign object on the belt conveyor in this thesis.According to the various foreign object pictures collected in the actual production environment,the original picture is expanded by six times using the image enhancement algorithm,and the common non-coal foreign object data set on the belt conveyor is established,and the data set is divided according to the PASCAL VOC2012 format.First,use the K-means algorithm to cluster the dimensions of common foreign object to obtain anchor sizes that are closer to the real non-coal foreign object,and then streamline the YOLOv3 feature extraction network and add 104×104 scale detection,keeping the detection time basically at the same time,it improves its recognition accuracy of foreign object.Finally,the classic target detection model Faster R-CNN,SSD,YOLOv3 and the improved YOLOv3 model are tested on the non-coal foreign object data set.The better detection effect proves that the improved YOLOv3 model is more suitable for detecting non-coal foreign object on the belt conveyor.In order to meet the engineering requirements of non-coal foreign object detection on belt conveyors,in this thesis,a non-coal foreign object detection system is developed based on the improved YOLOv3 model,which realizes real-time monitoring and noncoal object detection of coal-loaded belt conveyors.When the system detects non-coal foreign object,the detection result and foreign object information is automatically saved,the sound and light alarm for warning by belt conveyor control system,prompting the relevant staff to deal with the non-coal object in time.