Research and Application of a Lightweight Real-time Human Posture Detection Model

doi:10.16182/j.issn1004731x.joss.20-FZ0308

Abstract

Abstract: The traditional OpenPose model has good accuracy but slow speed in human posture detection. In order to accelerate the detection speed and reduce the model on condition of the detection precision, based on the traditional OpenPose model, the residual network with second-order term fusion is used to extract the low-level features, the weights of the trained model are pruned by the L1 norm weight, and an improved OpenPose model is proposed. Experiments show that when the detection accuracy is approximately equal to original model, the model size reduces to about 8%, the parameters reduces by nearly 83%, and the detection speed increases by 5 times. The improved OpenPose model is applied to the physical fitness test of sit-ups, and the results show that the detection accuracy of the model can reach 97%.

Key words: human pose detection, OpenPose model, Residual network system, weight pruning

CLC Number:

TP273
TH89

Zhu Hongkun, Yin Jiawei, Feng Wenyu, Hua Liang, Fei Minrui, Zhang Kun. Research and Application of a Lightweight Real-time Human Posture Detection Model[J]. Journal of System Simulation, 2020, 32(11): 2155-2165.

References

[1] Fang H,Xie S,Tai Y W,et al.RMPE:Regional Multi-person Pose Estimation [C].The 2017 IEEE International Conference on Computer Vision(ICCV),Venice.Italy:IEEE Press,2017:2353-2362.
[2] He K,Gkioxari G,Dollar P,et al.Mask R-CNN [C].The 2017 IEEE International Conference on Computer Vision(ICCV).Venice,Italy:IEEE Press,2017:2980-2988.
[3] Cao Z,Simon T,Wei S,et al.Realtime Multi-person 2D Pose Estimation Using Part Affinity Fields[C].Computer vision and pattern recognition(CVPR).Puerto Rico,USA:IEEE Press,2017:1302-1310.
[4] Wei S,Ramakrishna V,Kanade T,et al.Convolutional Pose Machines[C].Computer Vision and Pattern Recognition(CVPR).Las Vegas,USA:IEEE Press,2016:4724-4732.
[5] 周意乔,徐昱琳.基于双向LSTM的复杂环境下实时人体姿势识别[J].仪器仪表学报,2020,41(3):192-201.
Zhou Yiqiao,Xu Yulin.Real-time Human Posture Recognition in Complex Environment Based on Bidirectional LSTM[J].Chinese Journal of Scientific Instrument,2020,41(3):192-201.
[6] 周德良.OpenPose结合BP网络的行为检测方法[J].贵州大学学报(自然科学版),2020,37(3):87-92.
Zhou Deliang.Behavior Detection Method of OpenPose Combined With BP Network[J].Journal of Guizhou University(Natural Science Edition),2020,37(3):87-92.
[7] 郭园,郭晨旭,时新,等.基于OpenPose学习坐姿分析的桌椅人机适应性研究[J].林业工程学报,2020,5(2):179-185.
Guo Yuan,Guo Chenxu,Shi Xin,et al.Study on Human-machine Adaptability of Tables and Chairs Based on OpenPose Learning Sitting Posture Analysis[J].Journal of Forestry Engineering,2020,5(2):179-185.
[8] 彭秋萍,万莉莉,孙梦圆,等.基于机器视觉的扶梯自动急停系统[J].计算机系统应用,2020,29(3):87-92.
Peng Qiuping,Wan Lili,Sun Mengyuan,et al.Automatic Emergency Stop System of Escalator Based on Machine Vision[J].Application of Computer System,2020,29(3):87-92.
[9] Szegedy C,Liu W,Jia Y,et al.Going deeper with convolutions[C].The IEEE Conference on Computer Vision and Pattern Recognition(ICCV).Boston,USA:IEEE Press,2015:1-9.
[10] Costillareyes O,Verarodriguez R,Scully P J,et al.Analysis of Spatio-Temporal Representations for Robust Footstep Recognition with Deep Residual Neural Networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence(S0162-8828),2019,41(2):285-296.
[11] 刘望,孙金玉,马世伟.基于时空特征金字塔网络的动作时序检测方法[J].系统仿真学报,2019,31(11):2382-2387.
Liu Wang,Sun Jinyu,Ma Shiwei.A Temporal Action Detection Algorithm Based on Spatio-Temporal Feature Pyramid Network[J].Journal of System Simulation,2019,31(11):2382-2387.
[12] Liao Z,Wang J,Zhang S,et al.Minimizing Movement for Target Coverage and Network Connectivity in Mobile Sensor Networks[J].IEEE Transactions on Parallel and Distributed Systems(S1045-9219),2015,26(7):1971-1983.
[13] He K,Zhang X,Ren S,et al.Deep Residual Learning for Image Recognition[C].Computer Vision and Pattern Recognition(CVPR).Las Vegas,USA:IEEE Press,2016:770-778.
[14] Wang Y,Xie L,Liu C,et al.SORT:Second-Order Response Transform for Visual Recognition[C].International Conference on Computer Vision(ICCV).Venice,Italy:IEEE Press,2017:1368-1377.
[15] 石敏,姚瀚钦,李淳芃,等.基于深度Alignment网络的足部测量[J].系统仿真学报,2020,32(7):1267-1278.
Shi Min,Yao Hanqin,Li Chunpeng,et al.Foot Measurement Based on Deep Alignment Network[J].Journal of System Simulation,2020,32(7):1267-1278.
[16] Huang G,Liu S,Der Maaten L V,et al.CondenseNet:An Efficient DenseNet Using Learned Group Convolutions [C].Computer vision and pattern recognition(CVPR).Salt Lake City,USA:IEEE Press,2018:2752-2761.
[17] Zhou Z,Zhou W,Li H,et al.Online Filter Clustering and Pruning for Efficient Convnets[C].International conference on image processing(ICIP).Athens,Greece:IEEE Press,2018:11-15.
[18] 刘小雍,熊中刚,阎昌国.逼近误差的范数最小化的区间模糊模型建模[J].系统仿真学报,2018,30(3):1203-1209.
Liu Xiaoyong,Xiong Zhonggang,Yan Changguo.Interval Fuzzy Modeling Based on Minimizing-norm on Approximation Error[J].Journal of Forestry Engineering,2018,30(3):1203-1209.
[19] 徐嘉荟.基于模型剪枝的神经网络压缩技术研究[J].信息通信,2019(12):165-167.
Xu Jiahui.Research on Neural Network Compression Technology Based on Model Pruning[J].Information & Communications,2019(12):165-167.
[20] Duan P,Wang T,Cui M,et al.Multi-person Pose Estimation Based on a Deep Convolutional Neural Network[J].Journal of Visual Communication and Image Representation(S1047-3203),2019,61:245-252.