智能计算机与应用2025，Vol.15Issue(1) ：158-164.DOI:10.20169/j.issn.2095-2163.250124

基于结构化剪枝的矿区地质灾害检测算法

Geological hazard detection algorithm in mining area based on structured pruning

刘毅 高海海 韩英杰 张文杰 李鹏越
智能计算机与应用2025，Vol.15Issue(1) ：158-164.DOI:10.20169/j.issn.2095-2163.250124
✕
  • 维普
  • 万方数据

基于结构化剪枝的矿区地质灾害检测算法

Geological hazard detection algorithm in mining area based on structured pruning

刘毅 1高海海 1韩英杰 1张文杰 2李鹏越2
扫码查看

作者信息

  • 1. 华晋焦煤有限责任公司,山西 吕梁 033000
  • 2. 太原理工大学 电气与动力工程学院,太原 030024
  • 折叠

摘要

本文提出基于YOLOv5s模型的结构化剪枝目标检测算法,解决矿区无人机巡检中常规算法过大、参数多、难以部署的问题.通过遍历网络中的BN层,对γ进行排序,并设定全局阈值评估通道重要性,剔除低于阈值的通道.实验结果显示,相较于YOLOv5s,该算法模型减小 52.9%,检测时间降低 18.1%,平均精度仅下降 1.5%.

Abstract

In this paper,a structured pruning target detection algorithm based on YOLOv5s model is proposed to solve the problems of excessive size,many parameters and difficult deployment of conventional algorithms in UAV inspection in mining areas.By traversing the BN layer in the network,the γ are sorted and the global threshold is set to evaluate the importance of the channel and exclude channels below the threshold.The experimental results show that compared with YOLOv5s,the algorithm model is reduced by 52.9%,the detection time is reduced by 18.1%,and the average accuracy is only reduced by 1.5%.

关键词

矿区地质灾害/YOLOv5s/目标检测/结构化剪枝

Key words

Geological hazards in mining areas/YOLOv5s/object detection/structured pruning

引用本文复制引用

出版年

2025
智能计算机与应用
哈尔滨工业大学

智能计算机与应用

影响因子：0.357
ISSN：2095-2163
段落导航相关论文