| 基于深度图像的人体关节点定位方法 |
| |
| 引用本文： | 吕洁,刘亚洲,韩庆龙,杜晶.基于深度图像的人体关节点定位方法[J].海军航空工程学院学报,2016,31(5):538-546. |
| |
| 作者姓名： | 吕洁 刘亚洲 韩庆龙 杜晶 |
| |
| 作者单位： | 海军航空工程学院基础部,山东烟台 264001,南京理工大学计算机科学与工程学院,南京 210094,海军航空工程学院基础部,山东烟台 264001,海军航空工程学院基础部,山东烟台 264001 |
| |
| 摘 要： | 提出了一个基于深度图像的人体关节点定位的方法:首先将图像中的人体区域分割出来,然后利用随机森林分类器对逐个像素点进行分类,得到身体的各个部件并寻找关节点的位置。通过实验发现,本方法准确性较高并具有一定实时性。分类的准确率为 68%,相较 Kinect技术(40%)达到了较高的分类水平。预测人体关节点位置的平均时间为每帧 150ms,符合实用性要求。
|
| 关 键 词： | 深度图像 随机森林 深度梯度特征 关节位置定位 |
Method of Locationg Human Body Joints Based on Depth-images |
| |
| LYU Jie,LIU Yazhou,HAN Qinglong and DU Jing.Method of Locationg Human Body Joints Based on Depth-images[J].Journal of Naval Aeronautical Engineering Institute,2016,31(5):538-546. |
| |
| Authors: | LYU Jie LIU Yazhou HAN Qinglong and DU Jing |
| |
| Institution: | Department of Basic Sciences, NAAU, Yantai Shandong 264001, China,Department of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing 210094, China,Department of Basic Sciences, NAAU, Yantai Shandong 264001, China and Department of Basic Sciences, NAAU, Yantai Shandong 264001, China |
| |
| Abstract: | In this paper, a human body part recognition system which was based on depth image learning was proposed.Firstly, segmentation was made between background and people. Secondly, random forest was used to learn the classifierfor separating each pixel. Finally, the algorithm of mean-shift was used to find the position of the joint point of humanbody and optimize it. The result of the system was both more accurate and real-time. The precision rate for this methodwas 68%, which compared to Kinect, had reached at a high level. The average time to recognize joint point was 150ms/f.All these proved its practicability. |
| |
| Keywords: | depth images random forest depth gradient feature joint position proposals |
|
| 点击此处可从《海军航空工程学院学报》浏览原始摘要信息 |
| 点击此处可从《海军航空工程学院学报》下载免费的PDF全文 |
|
