首页 > 论文 > 激光与光电子学进展 > 55卷 > 6期(pp:61009--1)

基于时空方向主成分直方图的人体行为识别

Action Recognition Based on Histogram of Spatio-Temporal Oriented Principal Components

  • 摘要
  • 论文信息
  • 参考文献
  • 被引情况
  • PDF全文
分享:

摘要

为解决由视角、尺度等变化造成的行为类内差别大的问题, 提出一种基于时空方向主成分直方图(HSTOPC)的人体行为识别方法。首先, 将深度图序列转换为三维(3D)点云序列, 对此序列采用新颖的图像预处理方法获得新的深度图序列, 即在空间和时间维度上对采样获得的深度图序列进行限制, 除去一些动作信息量较少的序列和空间, 从而降低输入数据的冗余减少空间尺度变化的影响;其次, 为了解决帧间关联性较弱的问题, 采用时空方向主成分方法描述新的点云序列, 获得3D点云序列中每点的方向特征;然后, 对3D点云序列中所有方向特征进行多层时域重叠分割, 获得HSTOPC特征描述子;最后, 采用支持向量机分类器进行训练、测试。在3个标准数据库上的实验结果表明, 提出的HSTOPC特征描述子对噪声、运动速度变化、视角变化和时域不对齐具有很好的稳健性, 可以显著提高人体行为识别的准确率。

Abstract

In order to solve the problem of inter-class difference caused by the angle of view and scale change, we propose a method based on histogram of spatio-temporal oriented principal components of three-dimensional (3D) point clouds for action recognition. Firstly, the depth sequences are converted into 3D point clouds sequences. Then, we use a novel image preprocessing method to get new depth sequences. Namely, the sampled depth sequences are limited in spatio-temporal dimension to remove areas with less information, and reduce the redundancy of the input data and the influence of space scale change In order to solve the problem of weak correlation between frames, we adopt histogram of spatio-temporal oriented principal components (HSTOPC) method to describe 3D point clouds sequences and obtain the direction of each point of the 3D point clouds in sequences. For all direction of 3D point clouds in sequences, we use multilayer overlapping segmentation method to obtain HSTOPC descriptor. Finally, we use the support vector machine classifier for training and test. Experimental results on three human action recognition datasets show that the proposed HSTOPC feature descriptor has better robust for noise, rate variations, view change and temporal misalignment, and is able to improve the accuracy of human behavior recognition significantly.

Newport宣传-MKS新实验室计划
补充资料

中图分类号:TP391

DOI:10.3788/lop55.061009

所属栏目:图像处理

基金项目:国家自然科学基金(61362030,61201429)、中国博士后科学基金(2015M571720,2016M606360)、江苏省博士后科学基金(1601416C)、公安部技术研究计划(2014JSYJB007)

收稿日期:2017-11-28

修改稿日期:2017-12-28

网络出版日期:--

作者单位    点击查看

徐海洋:江南大学物联网工程学院, 江苏 无锡 214122
孔军:江南大学物联网工程学院, 江苏 无锡 214122新疆大学电气工程学院, 新疆 乌鲁木齐 830047
蒋敏:江南大学物联网工程学院, 江苏 无锡 214122
昝宝锋:江南大学物联网工程学院, 江苏 无锡 214122

联系人作者:孔军(kongjun@jiangnan.edu.cn)

备注:徐海洋(1992-), 男, 硕士研究生, 主要从事人体行为识别、机器学习等方面的研究。E-mail: 1041922650@qq.com

【1】Dong K. Human action recognition based on Kinect[D]. Wuhan: Wuhan University of Science and Technology, 2015.
董珂. 基于Kinect的人体行为识别研究[D]. 武汉: 武汉科技大学, 2015.

【2】Cai J X, Feng G C, Tang X, et al. Human action recognition based on poses a dictionary[J]. Acta Optica Sinica, 2014, 34(12): 1215002.
蔡加欣, 冯国灿, 汤鑫, 等. 基于姿势字典学习的人体行为识别[J]. 光学学报, 2014, 34(12): 1215002.

【3】Rahmani H, Mahmood A, Huynh D Q, et al. Real time human action recognition using histograms of depth gradients and random decision forests[C]. 2014 IEEE Winter Conference on Application of Computer Vision, 2014: 626-633.

【4】Cai J X, Feng G C, Tang X, et al. Human action recognition based on the local contour and random forest[J]. Acta Optica Sinica, 2014, 34(10): 1015006.
蔡加欣, 冯国灿, 汤鑫, 等. 基于局部轮廓和随机森林的人体行为识别[J]. 光学学报, 2014, 34(10): 1015006.

【5】Tang S, Wang X, Lv X, et al. Histogram of oriented normal vectors for object recognition with a depth sensor[C]. Asian Conference on Computer Vision, 2013: 525-538.

【6】Huang C H. Human action recognition algorithm based on video[D]. Chengdu: University of Electronic Science and Technology of China, 2016.
黄成挥. 基于视频的人体行为识别算法研究[D]. 成都: 电子科技大学, 2016.

【7】Wang J. Human action recognition based on sample learning method[J]. Information Technology, 2016(7): 65-70.
王军. 基于多示例学习法的人体行为识别[J]. 信息技术, 2016(7): 65-70.

【8】Liu Z, Dong S D. Human action recognition based on the joint movement in the depth of video information[J]. Computer Applications and Software, 2017, 34(2): 189-192.
刘智, 董世都. 利用深度视频中的关节运动信息研究人体行为识别[J]. 计算机应用与软件, 2017, 34(2): 189-192.

【9】Zhang Y J, Wang H M, Fu X H, et al. Identification of steel plate damage position based on particle swarm support vector machine[J]. Chinese Journal of Lasers, 2017, 44(10): 1006006.
张燕君, 王会敏, 付兴虎, 等. 基于粒子群支持向量机的钢板损伤位置识别[J]. 中国激光, 2017, 44(10): 1006006.

【10】Liu F,Sheng T S,Ma X X.Convolutional neural network based multi-band ship target recognition with feature fusion[J]. Acta Optica Sinica, 2017, 37(10): 1015002.
刘峰, 沈同圣, 马新星. 特征融合的卷积神经网络多波段舰船目标识别[J]. 光学学报, 2017, 37(10): 1015002.

【11】Bi L H, Liu Y C. Plant leaf image recognition based on improved neural network algorithm[J]. Laser & Optoelectronics Progress, 2017, 54(12): 121102.
毕立恒, 刘云潺. 基于改进神经网络算法的植物叶片图像识别研究[J]. 激光与光电子学进展, 2017, 54(12): 121102.

【12】Rahmani H, Mahmood A, Du Q H, et al. HOPC: histogram of oriented principal components of 3D pointclouds for action recognition[C]. European Conference on Computer Vision, 2014: 742-757.

【13】Cao L. Face recognition and human motion recognition technology and application[M]. Beijing: Electronic Industry Press, 2015.
曹林. 人脸识别与人体动作识别技术及应用[M]. 北京: 电子工业出版社, 2015.

【14】Li W, Zhang Z, Liu Z. Action recognition based on a bag of 3D points[C]. IEEE Computer Society Conference on Computer Vision and Pattern Recognition-Workshops, 2010: 9-14.

【15】Ji X, Cheng J, Feng W. Spatio-temporal cuboid pyramid for action recognition using depth motion sequences[C]. Eighth International Conference on Advanced Computational Intelligence, IEEE, 2016: 208-213.

【16】Zhang C, Tian Y. Histogram of 3Dfacets: a depth descriptor for human action and hand gesture recognition[J]. Computer Vision & Image Understanding, 2015, 139: 29-39.

【17】Wang C, Wang Y, Yuille A L. An Approach to Pose-Based Action Recognition[C]. Computer Vision and Pattern Recognition IEEE, 2013: 915-922.

【18】Rahmani H, Du Q H, Mahmood A, et al. Discriminative human action classification using locality-constrained linear coding[J]. Pattern Recognition Letters, 2016, 72: 62-71.

【19】Kong Y, Sattar B, Fu Y. Hierarchical 3D kernel descriptors for action recognition using depth sequences[C]. IEEE International Conference on Automatic Face and Gesture Recognition, 2015: 1-6.

【20】Chen C, Jafari R, Kehtarnavaz N. Action Recognition from Depth Sequences Using Depth Motion Maps-based Local Binary Patterns[C]. IEEE Applications of Computer Vision, 2015:1092-1099.

【21】Oreifej O, Liu Z. HON4D: Histogram of Oriented 4D Normals for Activity Recognition from Depth Sequences[C]. Computer Vision and Pattern Recognition, IEEE, 2013: 716-723.

引用该论文

Xu Haiyang,Kong Jun,Jiang Min,Zan Baofeng. Action Recognition Based on Histogram of Spatio-Temporal Oriented Principal Components[J]. Laser & Optoelectronics Progress, 2018, 55(6): 061009

徐海洋,孔军,蒋敏,昝宝锋. 基于时空方向主成分直方图的人体行为识别[J]. 激光与光电子学进展, 2018, 55(6): 061009

您的浏览器不支持PDF插件,请使用最新的(Chrome/Fire Fox等)浏览器.或者您还可以点击此处下载该论文PDF