Research on Coal Pile Spontaneous Combustion Detection Algorithm Based on Improved YOLOv5

Aiming at the problems of low accuracy and difficulty in providing real-time images in the application of current fire detection methods in complex outdoor actual environments, this paper proposes a coal pile spontaneous combustion detection algorithm based on improved YOLOv5 model. Specifically, a Dynamic Efficient Channel Attention (DECA) mechanism and a Dual-channel Feature Fusion (DPFF) module were designed. The DECA module enhances the model’s ability to process complex smoke features, and the DPFF module enhances the detection performance of small-scale smoke features by parallel processing paths. In this paper, the data of the yard environment are collected and processed to construct a high-quality image dataset for model training and testing. Experimental results show that our improved YOLOv5 algorithm is significantly better than other algorithms in terms of accuracy, recall rate and mAP@0.5, especially when detecting subtle and fuzzy smoke flow in coal piles. The importance of DECA and DPFF modules to improve the performance of early spontaneous combustion detection of coal piles was further verified in ablation experiments. This study not only significantly improves the accuracy and efficiency of fire early warning, but also provides new directions and insights for the application of deep learning technology in the broader field of disaster prevention and reduction. Early warning, but also provides new directions and insights for the application of deep learning technology in the broader field of disaster prevention and reduction.