摘要
煤矿井巷道存在着环境复杂、能见度低的特点,实时、准确地检测出巷道内的人员及位置是实现煤矿透明开采的前提条件。针对井下巷道内人员识别、定位的问题,提出了基于机器视觉技术的解决方案。通过巷道人员定位数据集训练模型,利用SE (Squeeze and Excitation)注意力机制对原始YOLOv5模型做出改进,得到SE-YOLOv5井下人员定位模型,并通过坐标转化求解出相对的三维坐标。试验结果表明,煤矿巷道人员定位模型能够准确地识别巷道内人员的三维坐标,检测精确率超过97.4%。
Due to the complex environment and low visibility of the coal mine roadway,real-time and accurate detection of the personnel and their locations in the roadway is a prerequisite for realizing transparent mining in coal mines.In this paper,aiming at the problems of personnel identification and positioning in underground roadways,a solution based on machine vision technology was proposed.Through the training model of roadway personnel positioning dataset,the original YOLOv5 model was improved by using SE(Squeeze and Excitation)attention mechanism,and the SE-YOLOv5 underground personnel positioning model was obtained.Then,the relative 3D coordinates were solved by coordinate transformation.The test results showed that the personnel positioning model of coal mine roadway could accurately identify the 3D coordinates of the personnel in the roadway,and the detection accuracy was more than 97.4%.
作者
王端
刘世平
王利军
WANG Duan;LIU Shiping;WANG Lijun(Wulanmulun Coal Mine,CHN Energy Shendong Coal Company,Ordos 017200,Inner Mongolia,China)
出处
《矿山机械》
2024年第1期56-60,共5页
Mining & Processing Equipment
作者简介
王端,男,1991年生,本科,工程师,主要从事智能化矿山建设网络数据维护工作。
