Automatic recognition of tramp metal in mining belt conveyors

LIU Huijun; LIU Keyi; GE Zhiqiang

doi:10.11799/ce202512022

Abstract：To improve image recognition accuracy,a tramp metal recognition method for mining belt conveyors based on machine vision and coal flow height is proposed.A binocular vision system is used to acquire images of tramp metal on the belt conveyor.Distortion is corrected using a planar template method to enhance image quality.The DA-GANomaly model in machine vision is utilized to generate a tramp metal recognition model.Based on the inferred coal flow height value,the model learns the feature distribution of normal coal flow images and identifies the difference between images containing tramp metal and the normal image feature distribution,thereby determining the presence of tramp metal.The CIoU loss function is selected,and an anomaly score is assigned to assist the model in judging whether an image contains tramp metal.Based on the anomaly score value,automatic recognition of tramp metal on mining belt conveyors is achieved.The results show that the proposed method has high recognition accuracy and efficiency,achieving the highest number of successful recognitions among the tested models.It can quickly and accurately determine whether an image contains tramp metal,enabling the recognition of tramp metal on the conveyor.