In recent years,coal mine safety has always been a key area of concern,and various technologies continuously provide better guarantees for coal mine production safety.Traditional video surveillance systems rely on manpower and suffer from blind spots and low efficiency.In order to improve the intelligent level of coal mine safety monitoring,a solution based on Hi3559A edge computing equipment and deep learning technology was proposed.This scheme generates training samples through Generative Adversarial Networks(GANs)and optimizes the YOLOv7 detection algorithm.A sliding window self attention mechanism(SwinTransformer),a dual feature extraction network(BiBranchBackbone),and a grid feature fusion module(GrideFuse)was introduced to improve detection and segmentation accuracy.To achieve precise detection and segmentation of open flames and smoke in complex scenes.The experimental results showed that the improved YOLOv7 algorithm had improved detection box mAP and segmentation mAP by 2.7％and 2.3％,respectively.The proposed intelligent coal mine safety warning and monitoring system has achieved significant results in practical applications,providing strong guarantees for the safety production of coal mining enterprises.
维普
万方数据