基于3D卷积神经网络的人体动作识别算法
|
摘要
由于人体动作的多样性、场景嘈杂、摄像机运动视角多变等特性,导致人体动作识别的难度增加。为此,基于3D卷积神经网络,提出一种新的人体动作识别算法。以连续的16帧视频为一组输入,采用视频图像的灰度、x方向梯度、y方向梯度、x方向光流、y方向光流做多通道处理,训练网络参数,经过5层3D卷积、5层3D池化增加提取特征中时间维度的动作信息,最终通过2层全连接与softmax分类器得到识别分类结果。在UCF101数据库上进行实验,结果表明,相比iDT、P-CNN、LRCN算法,该算法具有较高的识别准确率,且运行速度更快。
Human action diversity,scene noise,the camera motion angle changes and other factors increase the difficulty of human action recognition. This paper proposes a human action recognition algorithm based on 3D convolution neural network. Firstly,successive 16 frames of the video are divided into a group as the input. Secondly,the input data is multichannel processed using the gray,gradient-x,gradient-y,optflow-x and optflow-y,w hich effectively trains the network parameters. Thirdly,the extracted features are obtained using 5-layer 3D convolution and 5-layer 3D pooling to increase time dimension information,Finally,the recognition results are obtained by two full connection layers and the softmax classifier. Experiment is made on the UCF101 database,and the results show that compared with iDT,P-CNN,LRCN algorithms,the proposed algorithm has a higher accuracy of human action recognition and a faster running speed.
Human action diversity,scene noise,the camera motion angle changes and other factors increase the difficulty of human action recognition. This paper proposes a human action recognition algorithm based on 3D convolution neural network. Firstly,successive 16 frames of the video are divided into a group as the input. Secondly,the input data is multichannel processed using the gray,gradient-x,gradient-y,optflow-x and optflow-y,w hich effectively trains the network parameters. Thirdly,the extracted features are obtained using 5-layer 3D convolution and 5-layer 3D pooling to increase time dimension information,Finally,the recognition results are obtained by two full connection layers and the softmax classifier. Experiment is made on the UCF101 database,and the results show that compared with iDT,P-CNN,LRCN algorithms,the proposed algorithm has a higher accuracy of human action recognition and a faster running speed.
引文
[1]郑胤,陈权崎,章毓晋.深度学习及其在目标和行为识别中的新进展[J].中国图象图形学报,2014,19(2):175-184.
[2]徐勤军,吴镇扬.视频序列中的行为识别研究进展[J].电子测量与仪器学报,2014,28(4):343-351.
[3]雷庆,陈锻生,李绍滋.复杂场景下的人体行为识别研究新进展[J].计算机科学,2014,41(12):1-7.
[4]杜友田,陈峰,徐文立,等.基于视觉的人的运动识别综述[J].电子学报,2007,35(1):84-90.
[5]李岳云,许悦雷,马时平,等.深度卷积神经网络的显著性检测[J].中国图象图形学报,2016,21(1):53-59.
[6]DONAHUE J,HENDRICKS L A,GUADARRAMAS,et al.Long-term recurrent convolutional netw orks for visual recognition and description[J].IEEE Transactions on Pattern Analysis and M achine Intelligence,2014,39(4):677-691.
[7]李瑞峰,王亮亮,王珂.人体动作行为识别综述[J].模式识别与人工智能,2014,27(1):35-48.
[8]单言虎,张彰,黄凯奇.人的视觉行为识别研究回顾、现状及展望[J].计算机研究与发展,2016,53(1):93-112.
[9]KARPATHY A,TODERICI G,SHETTY S,et al.Largescale video classification w ith convolutional neural netw orks[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition(CVPR).Washington D.C.,USA:IEEE Press,2014:1725-1732.
[10]SIMONYAN K,ZISSERMAN A.Two-stream convolutional networks for action recognition in videos[J].Neural Information Processing Systems,2014(1):568-576.
[11]CHERON G,LAPTEV I,SCHMID C.P-CNN:Posedbased CNN features for action recognition[C]//Proceedings of IEEE Interational Conference on Computer Vision.Washington D.C.,USA:IEEE Press,2016:3218-3226.
[12]JI S,XU W,YANG M,et al.3D convolutional neural netw orks for human action recognition[J].IEEETransactions on Pattern Analysis and M achine Intelligence,2013,35(1):221-231.
[13]CHAQUET J M,CARMONA E J,FERNANDEZ-CABALLERO A.A survey of video datasets for human action and activity recognition[J].Computer Vision and Image Understanding,2013,117(6):633-659.
[14]SIMONYAN K,ZISSERMAN A.Very deep convolutional networks for large-scale image recognition[EB/OL].[2017-09-01].http://cn.arxiv.org/pdf/1409.1556v6.
[15]徐渊,许晓亮,李才年,等.结合SVM分类器与HOG特征提取的行人检测[J].计算机工程,2016,42(1):56-60.
[16]ZHANG B,WANG H.Encoding scale into fisher vector for human action recognition[C]//Proceedings of Visual Communications and Image Processing.Washington D.C.,USA:IEEE Press,2016:1-4.
[17]BEZAK P.Building recognition system based on deep learning[C]//Proceedings of International Conference on Artificial Intelligence and Pattern Recognition.Washington D.C.,USA:IEEE Press,2016:1-5.
[2]徐勤军,吴镇扬.视频序列中的行为识别研究进展[J].电子测量与仪器学报,2014,28(4):343-351.
[3]雷庆,陈锻生,李绍滋.复杂场景下的人体行为识别研究新进展[J].计算机科学,2014,41(12):1-7.
[4]杜友田,陈峰,徐文立,等.基于视觉的人的运动识别综述[J].电子学报,2007,35(1):84-90.
[5]李岳云,许悦雷,马时平,等.深度卷积神经网络的显著性检测[J].中国图象图形学报,2016,21(1):53-59.
[6]DONAHUE J,HENDRICKS L A,GUADARRAMAS,et al.Long-term recurrent convolutional netw orks for visual recognition and description[J].IEEE Transactions on Pattern Analysis and M achine Intelligence,2014,39(4):677-691.
[7]李瑞峰,王亮亮,王珂.人体动作行为识别综述[J].模式识别与人工智能,2014,27(1):35-48.
[8]单言虎,张彰,黄凯奇.人的视觉行为识别研究回顾、现状及展望[J].计算机研究与发展,2016,53(1):93-112.
[9]KARPATHY A,TODERICI G,SHETTY S,et al.Largescale video classification w ith convolutional neural netw orks[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition(CVPR).Washington D.C.,USA:IEEE Press,2014:1725-1732.
[10]SIMONYAN K,ZISSERMAN A.Two-stream convolutional networks for action recognition in videos[J].Neural Information Processing Systems,2014(1):568-576.
[11]CHERON G,LAPTEV I,SCHMID C.P-CNN:Posedbased CNN features for action recognition[C]//Proceedings of IEEE Interational Conference on Computer Vision.Washington D.C.,USA:IEEE Press,2016:3218-3226.
[12]JI S,XU W,YANG M,et al.3D convolutional neural netw orks for human action recognition[J].IEEETransactions on Pattern Analysis and M achine Intelligence,2013,35(1):221-231.
[13]CHAQUET J M,CARMONA E J,FERNANDEZ-CABALLERO A.A survey of video datasets for human action and activity recognition[J].Computer Vision and Image Understanding,2013,117(6):633-659.
[14]SIMONYAN K,ZISSERMAN A.Very deep convolutional networks for large-scale image recognition[EB/OL].[2017-09-01].http://cn.arxiv.org/pdf/1409.1556v6.
[15]徐渊,许晓亮,李才年,等.结合SVM分类器与HOG特征提取的行人检测[J].计算机工程,2016,42(1):56-60.
[16]ZHANG B,WANG H.Encoding scale into fisher vector for human action recognition[C]//Proceedings of Visual Communications and Image Processing.Washington D.C.,USA:IEEE Press,2016:1-4.
[17]BEZAK P.Building recognition system based on deep learning[C]//Proceedings of International Conference on Artificial Intelligence and Pattern Recognition.Washington D.C.,USA:IEEE Press,2016:1-5.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |