Scholarlink

<p>Against the backdrop of intelligent mining development, precise control of blasting fragmentation has become a critical factor in improving production efficiency and achieving energy conservation. However, existing fragmentation recognition methods rely largely on cumbersome manual operations, making it difficult to efficiently and accurately evaluate blasting performance. This study proposes an intelligent recognition method for blasting fragmentation to address limitations such as limited fragment samples, complex operational procedures, and low segmentation quality. A binocular camera is fixedly deployed to automatically capture images of rock fragments transported by passing trucks, and a You Only Look Once version 8 (YOLOv8) instance segmentation model is constructed to identify the fragments loaded on the trucks. The accuracy of fragment recognition and size estimation under dynamic conditions is validated through laboratory dynamic scanning experiments and field tests conducted in an open-pit mine. Experimental results demonstrate that the YOLOv8 segmentation model achieved a rock fragment recognition accuracy of 94.4%, with 94.12% of fragment size estimations exhibiting an accuracy exceeding 90%. To address the challenges posed by shadows and environmental disturbances under field conditions, a dynamic region of interest (DROI) method is proposed to accurately localize fragmented rocks within the truck bed. Comparative analysis against conventional image processing software reveals that the proposed method exhibits higher accuracy in fragment recognition, along with strong adaptability.</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

Blasting fragmentation recognition technology for intelligent mining: A synergistic application of deep learning and binocular vision

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月9日

查看全文
  • Yuanquan Xu,
  • Ming Tao,
  • Yulong Liu,
  • Lei Huang,
  • Gongliang Xiang,
  • Xibing Li

摘要

Against the backdrop of intelligent mining development, precise control of blasting fragmentation has become a critical factor in improving production efficiency and achieving energy conservation. However, existing fragmentation recognition methods rely largely on cumbersome manual operations, making it difficult to efficiently and accurately evaluate blasting performance. This study proposes an intelligent recognition method for blasting fragmentation to address limitations such as limited fragment samples, complex operational procedures, and low segmentation quality. A binocular camera is fixedly deployed to automatically capture images of rock fragments transported by passing trucks, and a You Only Look Once version 8 (YOLOv8) instance segmentation model is constructed to identify the fragments loaded on the trucks. The accuracy of fragment recognition and size estimation under dynamic conditions is validated through laboratory dynamic scanning experiments and field tests conducted in an open-pit mine. Experimental results demonstrate that the YOLOv8 segmentation model achieved a rock fragment recognition accuracy of 94.4%, with 94.12% of fragment size estimations exhibiting an accuracy exceeding 90%. To address the challenges posed by shadows and environmental disturbances under field conditions, a dynamic region of interest (DROI) method is proposed to accurately localize fragmented rocks within the truck bed. Comparative analysis against conventional image processing software reveals that the proposed method exhibits higher accuracy in fragment recognition, along with strong adaptability.