基于光流約束自編碼器的動(dòng)作識(shí)別
發(fā)布時(shí)間：2019-01-01 16:18
【摘要】：為了改進(jìn)特征學(xué)習(xí)在提取目標(biāo)運(yùn)動(dòng)方向及運(yùn)動(dòng)幅度等方面的能力,提高動(dòng)作識(shí)別精度,提出一種基于光流約束自編碼器的動(dòng)作特征學(xué)習(xí)算法.該算法是一種基于單層正則化自編碼器的無(wú)監(jiān)督特征學(xué)習(xí)算法,使用神經(jīng)網(wǎng)絡(luò)重構(gòu)視頻像素并將對(duì)應(yīng)的運(yùn)動(dòng)光流作為正則化項(xiàng).該神經(jīng)網(wǎng)絡(luò)在學(xué)習(xí)動(dòng)作外觀信息的同時(shí)能夠編碼物體的運(yùn)動(dòng)信息,生成聯(lián)合編碼動(dòng)作特征.在多個(gè)標(biāo)準(zhǔn)動(dòng)作數(shù)據(jù)集上的實(shí)驗(yàn)結(jié)果表明,光流約束自編碼器能有效提取目標(biāo)的運(yùn)動(dòng)部分,增加動(dòng)作特征的判別能力,在相同的動(dòng)作識(shí)別框架下該算法超越了經(jīng)典的單層動(dòng)作特征學(xué)習(xí)算法.
[Abstract]:In order to improve the ability of feature learning to extract the moving direction and amplitude of the target, and to improve the accuracy of motion recognition, an action feature learning algorithm based on optical flow constraint self-encoder is proposed. The algorithm is an unsupervised feature learning algorithm based on single-layer regularized self-encoder. Neural network is used to reconstruct the video pixels and the corresponding moving optical flow is used as the regularization term. The neural network can encode the motion information of objects while learning action appearance information and generate joint coded action features. The experimental results on several standard action data sets show that the optical flow constrained self-encoder can effectively extract the moving parts of the target and increase the ability to distinguish the motion features. Under the same framework of motion recognition, this algorithm surpasses the classical single-layer action feature learning algorithm.
【作者單位】： 東南大學(xué)自動(dòng)化學(xué)院;中國(guó)電科集團(tuán)28所;
【基金】：國(guó)家自然科學(xué)基金資助項(xiàng)目(61402426)
【分類號(hào)】：TP181;TP391.41
[Abstract]:In order to improve the ability of feature learning to extract the moving direction and amplitude of the target, and to improve the accuracy of motion recognition, an action feature learning algorithm based on optical flow constraint self-encoder is proposed. The algorithm is an unsupervised feature learning algorithm based on single-layer regularized self-encoder. Neural network is used to reconstruct the video pixels and the corresponding moving optical flow is used as the regularization term. The neural network can encode the motion information of objects while learning action appearance information and generate joint coded action features. The experimental results on several standard action data sets show that the optical flow constrained self-encoder can effectively extract the moving parts of the target and increase the ability to distinguish the motion features. Under the same framework of motion recognition, this algorithm surpasses the classical single-layer action feature learning algorithm.
【作者單位】： 東南大學(xué)自動(dòng)化學(xué)院;中國(guó)電科集團(tuán)28所;
【基金】：國(guó)家自然科學(xué)基金資助項(xiàng)目(61402426)
【分類號(hào)】：TP181;TP391.41
【相似文獻(xiàn)】
相關(guān)期刊論文 前10條
1 韓雷;王洪慶;林隱靜;;光流法在強(qiáng)對(duì)流天氣臨近預(yù)報(bào)中的應(yīng)用[J];北京大學(xué)學(xué)報(bào)(自然科學(xué)版);2008年05期
2 龔大墉;汪治華;楊數(shù)強(qiáng);劉巖;;基于目標(biāo)的局部約束光流分析[J];重慶工學(xué)院學(xué)報(bào)(自然科學(xué)版);2009年07期
3 嚴(yán)強(qiáng);黃增喜;曹麗萍;黃蓉剛;;光流法在機(jī)車安全行駛中的應(yīng)用[J];計(jì)算機(jī)應(yīng)用研究;2013年04期
4 馬鵬飛;楊金孝;;基于光流法的粒子圖像測(cè)速[J];科學(xué)技術(shù)與工程;2012年32期
5 張永亮;盧煥章;賀興華;謝耀華;;基于光流預(yù)測(cè)的直線對(duì)應(yīng)算法[J];信號(hào)處理;2010年05期
6 孫承志;熊田忠;吉順平;張家海;;基于差分的光流法在目標(biāo)檢測(cè)跟蹤中的應(yīng)用[J];機(jī)床與液壓;2010年14期
7 王，
本文編號(hào)：2397804
本文鏈接：http://sikaile.net/kejilunwen/zidonghuakongzhilunwen/2397804.html
最近更新
教材專著
