Improved Coal Gangue Recognition Detection Algorithm Based on YOLOv5s

Le résumé fourni par la source

To solve the problems of poor recognition and low speed of the coal gangue sorting task in the coal mining industry, a coal gangue recognition and detection algorithm based on improved YOLOv5s is proposed. The MobileNetV3 network is used as the backbone to reduce the network's calculation and parameter count. The SimAM attention module is integrated to further reduce the network's parameter count while improving the network's accuracy. At the output detection end of the network, a DyHead dynamic attention detection head is used to obtain more critical expression features in three dimensions: scale, space and task. By using the Alpha-IoU loss function, the detection network can more accurately locate the bounding box of the target object. The experimental results show that the improved algorithm achieves an average accuracy of 94.8%, an increase in detection speed to 109.38 FPS and a reduction in the number of parameters and computations to 59.57% and 46.88% respectively of YOLOv5s. The improved algorithm achieves the requirement for gangue detection.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.