光学技术, 2016, 42 (2): 146, 网络出版: 2016-04-01   

基于3D结构光传感器的老龄人异常行为检测方法

AAbnormal behavior detection for elderly based on 3D structure light sensor
罗坚 1,2,*唐琎 1赵鹏 1毛芳 1汪鹏 1
作者单位
1 中南大学 信息科学与工程学院, 湖南 长沙 410000
2 湖南信息职业技术学院 信息工程系, 湖南 长沙 410000
摘要
针对多视角下老龄人异常行为检测问题, 利用3D人体数据和多线性子空间分析方法, 从时间、视角和空间动作特征对其进行了研究。首先通过三维结构光传感器获取人体扫描点云数据, 并进行点云精简和人体表面重建。然后提取人体点云数据的表面曲率特征, 并将其映射到二维彩色图像中, 构成彩色动作特征图。通过提取特定时长内所有动作的彩色特征图, 生成基于曲率的彩色动作能量图模型, 并使用2D-PCA对彩色动作能量特征图进行降维。最后运用基于张量分析的多线性子空间分析方法, 对多视角下降维后的数据进行视角无关的特征提取, 并完成异常行为的分类和识别。实验结果表明, 该方法切实可行, 可将其应用于助老机器人和老龄人监护等相关领域。
Abstract
For the problems of multi-view elderly abnormal behavior detection, it is studied in real-time space, multi-view conditions and action features variations by 3D human body data and Multi-linear Subspace Analysis method. First, point cloud data reduction and 3D surface model reconstruction of the human body are achieved using 3D body scanning point cloud data captured by 3D structure light sensor. Then, the surface curvature features of human point cloud data are extracted and projected onto a 2D space to generate curvature based color motion feature images (CCMIs). By averaging the CCMIs of a motion cycle, a 2D curvature based color motion energy image (CCMEI) model is obtained. 2D principal component is used to reduce the dimensionality of CCMEI. Multi-linear Subspace Analysis method based on tensor decomposition is then introduced to extract view-invariant features from multi-view data after dimension reduction and conduct abnormal behavior classification and recognition. Experimental results demonstrate that the proposed method is feasible, and it can be applied to the service robots for the elderly, the aged care and related fields.

罗坚, 唐琎, 赵鹏, 毛芳, 汪鹏. 基于3D结构光传感器的老龄人异常行为检测方法[J]. 光学技术, 2016, 42(2): 146. LUO Jian, TANG Jin, ZHAO Peng, MAO Fang, WANG Peng. AAbnormal behavior detection for elderly based on 3D structure light sensor[J]. Optical Technique, 2016, 42(2): 146.

本文已被 1 篇论文引用
被引统计数据来源于中国光学期刊网
引用该论文: TXT   |   EndNote

相关论文

加载中...

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!