Abstract: During the coal mine belt conveyor process, the presence of foreign objects may cause safety risks such as conveying belt tearing or blockage.Aiming at the problems such as various foreign objects in conveyor belt materials, low efficiency of manual inspection and hardware limitations, a lightweight coal mine belt conveyor foreign object detection algorithm based on improved YOLOv8n was proposed. The original CSPDarkNet53 backbone network is lightweight improved by using GhostNetV2 network to reduce the parameters and calculation of the model. The SPPF module is optimized by integrating the ideas of global average pooling and global maximum pooling, and pays attention to the underlying information of the image under the harsh environmental impact of coal mines. The headC2f_CA module is designed to integrate the channel attention mechanism, so as to capture foreign object features in different scales and positions more effectively and enhance the expression of feature information.DIoU loss function is introduced to accurately reflect the similarity between anchor frame and prediction frame, and improve the detection accuracy of the model.The results show that the average accuracy of the improved model is 88.3%, which is 0.8% higher than that of the baseline model YOLOv8n. The number of parameters is reduced by 18.51%. The computational cost is reduced by 20.73%, and the model size is reduced by 15.87%. The model not only effectively alleviates the hardware limitations of edge devices, but also ensures the accuracy of coal mine safety monitoring.