Research on Lightweight Coal Gangue Recognition Algorithm Based on Improved YOLOv8n

Aiming at the problems that missed and false detections are prone to occur in traditional coal gangue detection algorithm, as well as the difficulty in achieving efficient deployment under the condition of limited computing resources, and others, a type of lightweight coal gangue recognition algorithm based on improved YOLOv8n is proposed. On the basis of YOLOv8n algorithm, the expansion residual structure DWR (Dilation-Wise Residual) is adopted to improve the C2f module of the backbone network part, and the accuracy of target identification is improved; The weighted bidirectional feature pyramid network BiFPN is introduced to improve detection capability and reduce parameter quantity; A type of lightweight detection head LSCSBD(Lightweight Shared Convolutional Separate Batch Normalization Detection Head) is designed to further reduce the computational complexity; Introducing Focaler-MPDIoU function, the detection accuracy of the model is improved. The test results show that the computational load, parameter quantity, and weight size of the improved YOLOv8n model compared to the original model is reduced by 26.8%, 47.9%, and 46.0%, respectively. The average mean accuracy value is improved by 1.4%, which reduces the problems of missed and false detections of coal gangue and improves the detection accuracy while ensuring the model lightweighting.