| 基于神经网络的姿态识别算法 |
| |
| 引用本文： | 张剑书,杨炼鑫,王浩然,樊英泽.基于神经网络的姿态识别算法[J].教育技术导刊,2009,19(11):33-36. |
| |
| 作者姓名： | 张剑书 杨炼鑫 王浩然 樊英泽 |
| |
| 作者单位： | 南京工程学院 计算机工程学院，江苏 南京211167 |
| |
| 基金项目： | 南京工程学院校级科研基金项目（QKJ201803）；江苏省高等学校大学生创新创业训练计划项目（201911276050Y） |
| |
| 摘 要： | 公共场所视频监控网络部署日益完善，智能视频监控技术在安防、交通等领域作用越来越大。针对视频监控数据中的人类目标，提出一种基于计算机视觉的姿态识别方法。通过YOLO算法和AlphaPose模型完成对视频中人类目标检测识别以及姿态估计，在此基础上分析人体关节之间的角度对姿态分类的影响，从中提取有效的分类特征，构造并训练5层神经网络模型，完成对站、坐、躺最常见3种姿态分类。实验结果表明，训练得到的神经网络模型准确率达到85%以上，识别速率大约为每秒30帧，在安防监控、检测人员摔倒、疾病报警等方面具有一定应用价值。
|
| 关 键 词： | 神经网络 目标检测识别 人体姿态估计 姿态分类 |
| 收稿时间： | 2020-06-25 |
Posture Recognition Algorithm Based on Neural Network |
| |
| ZHANG Jian-shu,YANG Lian-xin,WANG Hao-ran,FAN Ying-ze.Posture Recognition Algorithm Based on Neural Network[J].Introduction of Educational Technology,2009,19(11):33-36. |
| |
| Authors: | ZHANG Jian-shu YANG Lian-xin WANG Hao-ran FAN Ying-ze |
| |
| Institution: | School of Computer Engineering， Nanjing Institute of Technology， Nanjing 211167， China |
| |
| Abstract: | With the development of video surveillance network in public places， intelligent video surveillance technology plays an important role in security， transportation and other fields. In this paper， a computer vision based posture recognition method is proposed for human targets in video surveillance data. The detection， recognition and pose estimationof human targets in the video can be done through the YOLO algorithm and the AlphaPose model. On this basis， the influence of angles between human joints on posture classification is analyzed， and effective classification features are extracted from these angles. A five layer neural network model is constructed and trained to complete the classification of the three most common postures： standing， sitting and lying. The experimental results show that the accuracy rate of the trained neural network model can reach more than 85%， and the recognition rate is about 30 frames per second， and this method has a certain application value in security monitoring， staff falling and disease detection. |
| |
| Keywords: | neural network target detection and recognition human pose estimation pose classification |
|
| 点击此处可从《教育技术导刊》浏览原始摘要信息 |
| 点击此处可从《教育技术导刊》下载免费的PDF全文 |
