隧道工程一直是交通基础设施建设的重要组成部分。在隧道钻爆法施工中,装药环节对工程质量和施工效率起着关键作用。然而,传统的装药环节通常依赖于人工,存在效率低、危险性强的问题。本研究采用YOLOv5深度学习算法,更换骨干网络结构为S...隧道工程一直是交通基础设施建设的重要组成部分。在隧道钻爆法施工中,装药环节对工程质量和施工效率起着关键作用。然而,传统的装药环节通常依赖于人工,存在效率低、危险性强的问题。本研究采用YOLOv5深度学习算法,更换骨干网络结构为ShuffleNet V2,降低模型复杂程度;增加坐标注意力机制,提升检测准确性。通过自主收集和标记图像建立炮孔图像数据集。经训练后模型能够高效地检测和定位隧道炮孔,实现了准确性和速度的平衡。研究结果表明,改进的YOLOv5算法模型在满足检测精度的前提下,推理速度提升23%,模型大小降低54%。在隧道施工中具有巨大的潜力,有望在未来的建设中广泛应用。Tunneling is always an important part of the construction of transport infrastructure. In the construction of tunnel drilling and blasting methods, the loading link plays a crucial role in the project quality and construction efficiency. However, the traditional loading part usually relies on manual labour, which has the problems of low efficiency and danger. In this paper, the YOLOv5 deep learning algorithm is used to replace the backbone network structure with ShuffleNet V2 to reduce the model complexity;the coordinate attention mechanism is added to improve the detection accuracy. The cannon hole image dataset is established by autonomously collecting and labelling images. The trained model is able to detect and locate the tunnel gun holes efficiently, achieving a balance between accuracy and speed. The results show that the improved YOLOv5 algorithm model improves the inference speed by 23% and reduces the model size by 54% while satisfying the detection accuracy. It has great potential in tunnelling and is expected to be widely used in future construction.展开更多
文摘隧道工程一直是交通基础设施建设的重要组成部分。在隧道钻爆法施工中,装药环节对工程质量和施工效率起着关键作用。然而,传统的装药环节通常依赖于人工,存在效率低、危险性强的问题。本研究采用YOLOv5深度学习算法,更换骨干网络结构为ShuffleNet V2,降低模型复杂程度;增加坐标注意力机制,提升检测准确性。通过自主收集和标记图像建立炮孔图像数据集。经训练后模型能够高效地检测和定位隧道炮孔,实现了准确性和速度的平衡。研究结果表明,改进的YOLOv5算法模型在满足检测精度的前提下,推理速度提升23%,模型大小降低54%。在隧道施工中具有巨大的潜力,有望在未来的建设中广泛应用。Tunneling is always an important part of the construction of transport infrastructure. In the construction of tunnel drilling and blasting methods, the loading link plays a crucial role in the project quality and construction efficiency. However, the traditional loading part usually relies on manual labour, which has the problems of low efficiency and danger. In this paper, the YOLOv5 deep learning algorithm is used to replace the backbone network structure with ShuffleNet V2 to reduce the model complexity;the coordinate attention mechanism is added to improve the detection accuracy. The cannon hole image dataset is established by autonomously collecting and labelling images. The trained model is able to detect and locate the tunnel gun holes efficiently, achieving a balance between accuracy and speed. The results show that the improved YOLOv5 algorithm model improves the inference speed by 23% and reduces the model size by 54% while satisfying the detection accuracy. It has great potential in tunnelling and is expected to be widely used in future construction.
