| 人体姿态检测系统的设计与实现 |
| |
| 引用本文： | 陈博,李占文,杨永成,钟昊,宗孝鹏.人体姿态检测系统的设计与实现[J].铁路计算机应用,2021,30(4):56-60. |
| |
| 作者姓名： | 陈博 李占文 杨永成 钟昊 宗孝鹏 |
| |
| 作者单位： | 1.中国铁路兰州局集团有限公司 安全监察室，兰州 730030 |
| |
| 基金项目： | 中国铁路兰州局集团有限公司重点课题（lj-700_1000-20200623-2001） |
| |
| 摘 要： | 针对铁路岗位监控视频需要大量人力查看的问题,设计并实现了人体姿态检测系统,实现7×24 h不间断智能检测。运用人体姿态检测算法对人体关键点进行提取,构建特征向量,通过支持向量机分类算法得到姿态类别,通过滑动窗口法判断是否有异常行为,当发生异常时进行本地提醒和远程报警。在行车室和机车司机室2种环境下进行测试,结果表明,该方法对于人员异常行为的召回率和准确率均达到97%以上,具有实用价值,可以为铁路岗位管控提供有效方法。
|
| 关 键 词： | 人体姿态 特征向量 支持向量机 滑动窗口 岗位管控 |
| 收稿时间： | 2020-10-22 |
Design and implementation of human posture detection system |
| |
| Affiliation: | 1.Safety Supervision Office, China Railway Lanzhou Group Co. Ltd., Lanzhou 730030, China2.Jiaxun Feihong Intelligent Technology Institute, Beijing 100044, China |
| |
| Abstract: | In view of the problem that the railway post monitoring video needs a lot of manpower to view, a human posture detection system was designed and implemented to realize 7 × 24 h uninterrupted intelligent detection.This paper used the human posture detection algorithm to extract the key points of the human body, constructed the feature vector, obtained the posture category through the support vector machine classification algorithm, judged whether there was abnormal behavior through the sliding window method, and carried on the local reminder and remote alarm when there was abnormal. The system was tested in the railway operation room and locomotive cab. The results show that the recall rate and accuracy rate of abnormal behavior are above 97%, which has practical value and can provide an effective method for railway post management and control. |
| |
| Keywords: | |
|
| 点击此处可从《铁路计算机应用》浏览原始摘要信息 |
|
点击此处可从《铁路计算机应用》下载全文 |
|
