基于融合型深度学习的人体动态特征提取
|
摘要
为了提高对不同运动状态下人体动态特征的分析能力,提出了一种基于融合型深度学习的人体动态特征提取算法。采用图像亮点流形标注方法进行人体图像的动态特征采样,对动态图像采用RGB颜色特征分解方法进行灰度像素二值化拟合处理,采用多尺度小波分解方法实现行人的差异性特征提取,对所提取的人体动态特征量采用深度学习方法进行自适应分类处理,使用融合型卷积神经网络对分类后的动态特征量进行超分辨融合,实现了人体动态特征的优化提取。仿真结果表明,采用该方法进行人体动态特征提取的超分辨性较好,在时间开销和图像识别精度方面具有优越性。
In order to improve the ability of dynamic feature analysis of human body in different motion states, a human body dynamic feature extraction algorithm based on fusion depth learning is proposed. The dynamic feature sampling of human body image is carried out by using the image highlight manifold annotation method, and the gray pixel binary fitting processing is carried out by using the RGB color feature decomposition method for the dynamic image. The multi-scale wavelet decomposition method is used to extract the pedestrian differential features, and the depth learning method is used to self-adaptively classify the extracted human dynamic features. The fusion convolution neural network is used to perform super-resolution fusion of the classification dynamic feature to realize the optimal extraction of the human body dynamic feature. The simulation results show that this method has good super-resolution in extracting human dynamic features and has advantages in execution time overhead and image recognition accuracy.
In order to improve the ability of dynamic feature analysis of human body in different motion states, a human body dynamic feature extraction algorithm based on fusion depth learning is proposed. The dynamic feature sampling of human body image is carried out by using the image highlight manifold annotation method, and the gray pixel binary fitting processing is carried out by using the RGB color feature decomposition method for the dynamic image. The multi-scale wavelet decomposition method is used to extract the pedestrian differential features, and the depth learning method is used to self-adaptively classify the extracted human dynamic features. The fusion convolution neural network is used to perform super-resolution fusion of the classification dynamic feature to realize the optimal extraction of the human body dynamic feature. The simulation results show that this method has good super-resolution in extracting human dynamic features and has advantages in execution time overhead and image recognition accuracy.
引文
[1] 孙毅堂,宋慧慧,张开华,等.基于极深卷积神经网络的人脸超分辨率重建算法[J].计算机应用,2018,38(4):1141-1145.
[2] 傅天宇,金柳颀,雷震,等.基于关键点逐层重建的人脸图像超分辨率方法[J].信号处理,2016,32(7):834-841.
[3] ZHANG J Y,ZHAO H P,CHEN S.Face recognition based on weighted local binary pattern with adaptive threshold[J].Jounal of Electronics & Information Technology,2014,36(6):1327-1333.
[4] NASEEM I,TOGNERI R,BENNAMOUN M.Linear regression for face recognition[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2010,32(11):2106-2112.
[5] 李明爱,崔燕,杨金福,等.基于HHT和CSSD的多域融合自适应脑电特征提取方法[J].电子学报,2013, 41(12):2479-2486.
[6] GRIMM F,NAROS G,GHARABAGHI A.Closed-loop task difficulty adaptation during virtual reality reach-to-grasp training assisted with an exoskeleton for stroke rehabilitation[J].Frontiers in Neuroscience,2016(10):518.
[7] MUKAINO M,ONO T,SHINDO K,et al.Efficacy of brain-computer interface-driven neuromuscular electrical stimulation for chronic paresis after stroke[J].Journal of Rehabilitation Medicine,2014,46(4):378-382.
[8] FERRARA P,BIANCHI T,ALESSIA DR,et al.Image forgery localization via fine-grained analysis of CFA artifacts[J]. IEEE Transactions on Information Forensics and Security,2012,7(5):1566-1577.
[9] LONG T F,JIAO W L,HE G J,et al.Automatic line segment registration using Gaussian mixture model and expectation-maximization algorithm[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2014, 7(5): 1688-1699.
[10] LYU S W,PAN X Y,ZHANG X.Exposing region splicing forgeries with blind local noise estimation[J].International Journal of Computer Vision,2013,110(2):202-221.
[11] 宋颖超,罗海波,惠斌,等.尺度自适应暗通道先验去雾方法[J].红外与激光工程,2016,45(9):286-297.
[12] 郭丙华,岑志松.小波去噪和神经网络相融合的超分辨率图像重建[J].激光杂志,2016,37(2):61-64.
[2] 傅天宇,金柳颀,雷震,等.基于关键点逐层重建的人脸图像超分辨率方法[J].信号处理,2016,32(7):834-841.
[3] ZHANG J Y,ZHAO H P,CHEN S.Face recognition based on weighted local binary pattern with adaptive threshold[J].Jounal of Electronics & Information Technology,2014,36(6):1327-1333.
[4] NASEEM I,TOGNERI R,BENNAMOUN M.Linear regression for face recognition[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2010,32(11):2106-2112.
[5] 李明爱,崔燕,杨金福,等.基于HHT和CSSD的多域融合自适应脑电特征提取方法[J].电子学报,2013, 41(12):2479-2486.
[6] GRIMM F,NAROS G,GHARABAGHI A.Closed-loop task difficulty adaptation during virtual reality reach-to-grasp training assisted with an exoskeleton for stroke rehabilitation[J].Frontiers in Neuroscience,2016(10):518.
[7] MUKAINO M,ONO T,SHINDO K,et al.Efficacy of brain-computer interface-driven neuromuscular electrical stimulation for chronic paresis after stroke[J].Journal of Rehabilitation Medicine,2014,46(4):378-382.
[8] FERRARA P,BIANCHI T,ALESSIA DR,et al.Image forgery localization via fine-grained analysis of CFA artifacts[J]. IEEE Transactions on Information Forensics and Security,2012,7(5):1566-1577.
[9] LONG T F,JIAO W L,HE G J,et al.Automatic line segment registration using Gaussian mixture model and expectation-maximization algorithm[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2014, 7(5): 1688-1699.
[10] LYU S W,PAN X Y,ZHANG X.Exposing region splicing forgeries with blind local noise estimation[J].International Journal of Computer Vision,2013,110(2):202-221.
[11] 宋颖超,罗海波,惠斌,等.尺度自适应暗通道先验去雾方法[J].红外与激光工程,2016,45(9):286-297.
[12] 郭丙华,岑志松.小波去噪和神经网络相融合的超分辨率图像重建[J].激光杂志,2016,37(2):61-64.