Lightweight YOLOv8 Mine Target Detection Algorithm for Embedded Deployment

Abstract

Mine safety monitoring systems are usually deployed on embedded devices with limited computing, storage and energy resources. Traditional YOLOv8 models have large number of parameters and high computational complexity, which are difficult to realize real-time inference on these devices. In order to solve this problem, this study proposes a lightweight YOLOv8 mine target detection algorithm. The original PAN-FPN structure is replaced with BiFPN bidirectional feature pyramid to improve multi-scale feature reuse efficiency and reduce invalid computation through adaptive weighted fusion. A GCHead lightweight structure based on Group Convolution is introduced into the detection head to reduce parameter dimension and computational cost while maintaining detection accuracy. Experimental results show that compared with the original YOLOv8, the number of parameters and computational complexity of YOLOv8 are reduced by 46.5% and 37.8% respectively, while mAP@0.5 only decreases by 0.5%. It achieves the optimal balance between detection accuracy and lightweight, and can be efficiently deployed on mine embedded monitoring terminals. This research provides a useful reference for the application of target detection technology in resource-constrained environments.