Enhanced Ground Fissure Detection in Mining Areas Based on Visible–Infrared Image Fusion and YOLOv5

Intensive mining operations can lead to the formation of fissures caused by ground subsidence, which present significant threats to building stability, coal mine safety, and the ecological environment. Accurate and timely detection of these fissures is crucial for effective risk management and mitigation. This article proposes FisFusionYOLO, a novel method that integrates visible–infrared image fusion with a YOLOv5 deep learning network to enhance fissure detection accuracy and efficiency. By combining complementary information from visible and infrared images, the fusion strategy improves the representation of fissure features, which are then processed by the YOLOv5 network for precise and efficient object detection. A dataset collected from the Daliuta Mine in the Shendong Mining Area, using a uncrewed aerial vehicle (UAV) equipped with infrared and visible sensors, demonstrates the effectiveness of the FisFusionYOLO. The method achieves a mean average precision (mAP) score of 82.6%, surpassing those trained on visible and infrared image datasets. Furthermore, FisFusionYOLO exhibits superior generalization performance (77.1% mAP), compared to 24.2% for the visible image detector and 24.2% for the infrared image detector. A statistical analysis of fissure distribution and self-healing properties, based on the detection results, provides valuable insights for proactive risk mitigation. This approach offers a robust, automated solution for monitoring ground fissures in mining areas by integrating advanced image fusion techniques with deep learning. The proposed method can contribute to improved safety practices and environmental protection by enabling early detection and systematic assessment of fissure-related hazards.