中国安全科学学报 ›› 2019, Vol. 29 ›› Issue (1): 7-12.doi: 10.16265/j.cnki.issn1003-3033.2019.01.002

• 安全人体学 • 上一篇    下一篇

人工智能技术在矿工不安全行为识别中的融合应用

佟瑞鹏 教授, 张艳伟   

  1. 中国矿业大学(北京) 应急管理与安全工程学院,北京 100083
  • 收稿日期:2018-10-20 修回日期:2018-12-01 出版日期:2019-01-28 发布日期:2020-11-23
  • 作者简介:佟瑞鹏 (1977—),男,黑龙江穆棱人,博士,教授,主要从事行为安全与管理、风险建模与评估、公共安全与健康等方面的研究。E-mail:tongrp@cumtb.edu.cn。张艳伟 (1995—),男,内蒙古通辽人,硕士研究生,主要研究方向为行为安全与管理。E-mail:412572610@qq.com。
  • 基金资助:
    国家自然科学基金资助(51674268)。

Integration between artificial intelligence technologies for miners' unsafe behavior identification

TONG Ruipeng, ZHANG Yanwei   

  1. School of Emergency Management & Safety Engineering,China University of Mining and Technology (Beijing),Beijing 100083,China
  • Received:2018-10-20 Revised:2018-12-01 Online:2019-01-28 Published:2020-11-23

摘要: 为有效识别矿工不安全行为,预防煤矿安全事故,提出融合深度学习的计算机视觉、表示深度信息的深度图像、可穿戴传感器等人工智能识别技术的方法。基于以上3种方法在人体行为识别上的应用特点,运用主成分分析法(PCA)将3种识别技术提取的行为特征降维融合,通过支持向量机(SVM)对融合特征进行分类;以矿工跌倒行为数据为正样本,走路、坐下、弯腰、下蹲、躺下等5种日常行为数据作为负样本,分别利用3种人工智能识别方法以及融合方法对矿工跌倒行为进行识别检验。结果表明:经过融合后的识别方法对矿工跌倒行为的识别效果均高于其他3种人工智能识别方法。

关键词: 不安全行为, 智能识别, 主成分分析法(PCA), 行为特征, 跌倒检测

Abstract: A fusion method, for identifying unsafe behavior of miners, was worked out as a result of integration between three existing artificial intelligence identification methods including the computer vision based on depth learning, depth image representing depth information and wearable sensor. The method uses PCA to reduce the dimensions of the behavior features extracted by the three recognition techniques, and classifies the features by support vector machine (SVM). Data on miners' fall behavior were used as positive samples and data on five kinds of daily behavior including walking, sitting down, bending, squatting and lying down were used as negative samples. Three artificial intelligence identification methods and the fusion method were applied to identify the fall behavior of miners. The results show that the effectiveness of the fusion method in recognizing unsafe behavior is higher than that of the three artificial intelligence methods.

Key words: unsafe behavior, intelligent identification, principal component analysis (PCA), behavior characteristics, falling test

中图分类号: