[1]宫法明,马玉辉.基于时空双分支网络的人体动作识别研究[J].计算机技术与发展,2020,30(09):23-28.[doi:10. 3969 / j. issn. 1673-629X. 2020. 09. 005]
　GONG Fa-ming,MA Yu-hui.Research on Human Action Recognition Based on Space-time Double-branch Network[J].,2020,30(09):23-28.[doi:10. 3969 / j. issn. 1673-629X. 2020. 09. 005]
点击复制

基于时空双分支网络的人体动作识别研究()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
30
期数:
2020年09期
页码:
23-28
栏目:
智能、算法、系统工程
出版日期:
2020-09-10

文章信息/Info

Title:
Research on Human Action Recognition Based on Space-time Double-branch Network
文章编号:
1673-629X(2020)09-0023-06
作者:
宫法明; 马玉辉
中国石油大学(华东) 计算机科学与技术学院,山东 青岛 266580
Author(s):
GONG Fa-ming; MA Yu-hui
School of Computer Science and Technology,China University of Petroleum,Qingdao 266580,China
关键词:
人体动作识别; 关键点检测; 目标检测; 动作分类; 卷积姿态机; 深度学习
Keywords:
human action recognition; key point detection; object detection; action classification; convolutional pose machines; deep learning
分类号:
TP391
DOI:
10. 3969 / j. issn. 1673-629X. 2020. 09. 005
摘要:
常规的人体动作识别算法在单一特定的场景中效果较为突出,但在海洋钻井平台的实际工程场景中,易受管道遮挡和干扰,不能充分地利用视频的时序结构信息。 针对这些问题,提出了一种复杂场景下基于时空双分支网络的人体动作识别框架。 采用多规则区域提案标记算法将海水区域分离,将先验知识加入支持向量机分类器,提出后验判别准则以去除非人员目标,通过目标定位与检测算法分割出人员目标,利用卷积姿态机算法进行身体部位定位和关联程度分析以提取全部人体关键点信息,形成关键点序列;通过双分支网络对人体关键点轨迹和光流轨迹叠加融合,完成了人体动作的分类与识别。 实验结果表明,该方法实现了人体不可见关键点的检测和估计,免去了人工标注目标的繁杂工作,能够有效地解决海洋平台场景下的人体动作识别问题。
Abstract:
The conventional human action recognition algorithm is more effective in a single specific scene. However,it is vulnerable to pipeline occlusion and interference in the actual engineering scene of offshore oil platforms,and the timing structure information of the video cannot be fully utilized. Aiming at these problems,a human action recognition framework based on spatiotemporal double-branch network in complex scenes is proposed. The multi-rule regional proposal marker algorithm is used to separate the seawater regions,the prior knowledge is added to the SVM classifier,and the posterior discriminant criterion is proposed to remove the non-personnel targets.The target location and detection algorithm is used to segment the human target,and the convolutional pose machines algorithm is used to perform body part localization and correlation degree analysis to extract all key information of the human body to form a key point sequence. The double-branch network superimposes the human key point trajectory and the optical flow trajectory,and completes the classification and recognition of human action. The experiment shows that the proposed method realizes the detection and estimation of the invisible key points of the human body,eliminating the complicated work of manually marking the target,which can effectively solve the problem of human action recognition under the ocean platform scene.

相似文献/References:

[1]高 鹏,张 岩,唐新余,等.结合注意力机制的雷达多信号动作识别方法[J].计算机技术与发展,2023,33(01):157.[doi:10. 3969 / j. issn. 1673-629X. 2023. 01. 024]
　GAO Peng,ZHANG Yan,TANG Xin-yu,et al.Radar Multi-signal Action Recognition Method Based on Attention Mechanism[J].,2023,33(09):157.[doi:10. 3969 / j. issn. 1673-629X. 2023. 01. 024]
[2]朱联祥,牛文煜,仝文东,等.基于混合注意力机制的视频人体动作识别[J].计算机技术与发展,2023,33(09):105.[doi:10. 3969 / j. issn. 1673-629X. 2023. 09. 016]
　ZHU Lian-xiang,NIU Wen-yu,TONG Wen-dong,et al.Video Human Action Recognition Based on Hybrid Attention Mechanism[J].,2023,33(09):105.[doi:10. 3969 / j. issn. 1673-629X. 2023. 09. 016]

更新日期/Last Update: 2020-09-10