Scholarlink

<p>UAV-based aerial imagery plays a vital role in detecting surface cracks in mining-induced areas for geological disaster early warning and safe production. However, detection is challenged by small crack size, complex morphology, large scale variation, and uneven spatial distribution, further exacerbated by UAVs’ limited onboard computational capacity. To tackle these issues, we introduce an efficient and lightweight small-target detection model, namely YOLO-LSN, which is built upon the optimized YOLO architecture.Firstly, we introduce a Lightweight Dynamic Alignment Detection Head (LDADH) for multi-scale feature fusion, precise alignment, and dynamic receptive field adjustment, optimizing crack feature extraction. Secondly, the Small Object Feature Enhancement Pyramid (SOFEP) enhances detail representation of small cracks in complex backgrounds.Furthermore, we propose a weighted combination strategy of Normalized Wasserstein Distance (NWD) and IoU loss, balancing sensitivity to zero-overlap instances and robustness against scale deviations, thereby improving localization accuracy and generalization capability. Experiments show a 12% mAP@0.5 improvement and 17% reduction in parameters on a self-built mining crack dataset, with further validation on VisDrone2019 (mAP@0.5: 0.422, + 11.6%), Validating its effectiveness for UAV-based small-object detection, the model offers an efficient, reliable solution for geological hazard warning and mining safety.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Research on UAV aerial imagery detection algorithm for Mining-Induced surface cracks based on improved YOLOv10

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月17日

查看全文
  • Jiayong An,
  • Siyuan Dong,
  • Xuanli Wang,
  • Chenlei Li,
  • Wenpei Zhao

摘要

UAV-based aerial imagery plays a vital role in detecting surface cracks in mining-induced areas for geological disaster early warning and safe production. However, detection is challenged by small crack size, complex morphology, large scale variation, and uneven spatial distribution, further exacerbated by UAVs’ limited onboard computational capacity. To tackle these issues, we introduce an efficient and lightweight small-target detection model, namely YOLO-LSN, which is built upon the optimized YOLO architecture.Firstly, we introduce a Lightweight Dynamic Alignment Detection Head (LDADH) for multi-scale feature fusion, precise alignment, and dynamic receptive field adjustment, optimizing crack feature extraction. Secondly, the Small Object Feature Enhancement Pyramid (SOFEP) enhances detail representation of small cracks in complex backgrounds.Furthermore, we propose a weighted combination strategy of Normalized Wasserstein Distance (NWD) and IoU loss, balancing sensitivity to zero-overlap instances and robustness against scale deviations, thereby improving localization accuracy and generalization capability. Experiments show a 12% mAP@0.5 improvement and 17% reduction in parameters on a self-built mining crack dataset, with further validation on VisDrone2019 (mAP@0.5: 0.422, + 11.6%), Validating its effectiveness for UAV-based small-object detection, the model offers an efficient, reliable solution for geological hazard warning and mining safety.