摘要
为解决高铁建设大型装备作业时因为视觉盲区大而产生的人员安全问题,以高铁建设装备运行的典型场所,各工况下不同作业人员典型姿态图像为样本,采用Haar-like特征与Adaboost算法结合的训练方法,构建人员图像识别分类器,为提高分类器的准确性,在训练方法加入正样本自更新的方法以作改进。利用OpenCV分类器训练框架和Python语言,构建了高铁建设装备作业区域内人员的防入侵在线检测系统。试验和现场的运行验证了正样本自更新方法的可行性及系统的可靠性。
In order to solve the problem of personnel safety caused by the large visual blind area in the operation of large-scale equipment for high-speed railway construction,a classifier of personnel image recognition is constructed by combining Haarlike features with Adaboost algorithm,taking the typical places of high-speed railway construction equipment operation and the typical posture images of different operators under different working conditions as samples. In order to improve the accuracy of the classifier,a positive sample self-renewal method is added to the training method for improvement. Using OpenCV classifier training framework and Python language,an on-line intrusion detection system for personnel in the operation area of high-speed railway construction equipment is constructed. The feasibility of the positive sample self-renewal method and the reliability of the system are verified by the test and field operation.
作者
马利伟
周明阳
刘国宁
MA Li-wei;ZHOU Ming-yang;LIU Guo-ning(School of Mechanical Engineering,Zhengzhou University,He,nan Zhengzhou 450001,China)
出处
《机械设计与制造》
北大核心
2022年第2期182-187,共6页
Machinery Design & Manufacture
基金
国家重点研发计划项目(2018YFB1201403)
河南省产学研合作项目(182107000053)
赛尔网络下一代互联网技术创新项目(NGII20180702)。
