计算机工程与应用 ›› 2022, Vol. 58 ›› Issue (1): 190-196.DOI: 10.3778/j.issn.1002-8331.2007-0218

• 模式识别与人工智能 • 上一篇    下一篇

复杂姿态下的安全帽佩戴检测方法研究

王雨生,顾玉宛,庄丽华,石林,李宁,徐守坤   

  1. 常州大学 计算机与人工智能学院 阿里云大数据学院,江苏 常州 213164
  • 出版日期:2022-01-01 发布日期:2022-01-06

Research on Detection Method of Helmet Wearing in Complex Posture

WANG Yusheng, GU Yuwan, ZHUANG Lihua, SHI Lin, LI Ning, XU Shoukun   

  1. Aliyun School of Big Data, School of Computer and Artificial Intelligence, Changzhou University, Changzhou, Jiangsu 213164, China
  • Online:2022-01-01 Published:2022-01-06

摘要: 在施工人员复杂姿态下,现有安全帽佩戴检测方法存在检测难度大,精度不高的问题,提出一种基于头部识别的安全帽佩戴检测方法。通过肤色特征识别和头部检测获取头部区域,并进行交叉验证确定头部区域,使用YOLOv4目标检测网络识别安全帽,通过安全帽区域与头部区域的位置关系判断安全帽佩戴情况。最后,通过实验对比分析其他安全帽佩戴检测方法的性能,对安全帽佩戴检测方法进行总结并提出展望。

关键词: 目标检测, 安全帽识别, 头部识别, YOLOv4

Abstract: Existing helmet wearing detection methods have the problems of high detection difficulty and low accuracy in the case of complicated postures for construction workers. A helmet wearing detection method is proposed based on head recognition. Firstly, the head area is obtained through skin color feature recognition and head detection, which performs cross-validation to determine the head area. Secondly, YOLOv4 target detection network is adopted to identify the helmet and determine the positional relationship between the helmet area and the head area. Finally, the performance of other helmet wearing detection methods is analyzed through experiments, the helmet wearing detection methods are summarized and prospects are proposed.

Key words: target detection, helmet recognition, head recognition, YOLOv4