發(fā)布時(shí)間：2018-10-16 12:42

【摘要】：近年來(lái),高清視頻設(shè)備的推出使得基于行為識(shí)別技術(shù)的人工智能在智慧安全城市、智能家居和軍事安防等領(lǐng)域得以飛速發(fā)展。廣泛的應(yīng)用前景和經(jīng)濟(jì)價(jià)值讓行為分析與識(shí)別這一技術(shù)迅速成為計(jì)算機(jī)視覺(jué)領(lǐng)域的研究熱點(diǎn)。傳統(tǒng)的行為識(shí)別算法通常分為運(yùn)動(dòng)前景檢測(cè)、特征提取以及訓(xùn)練識(shí)別三個(gè)步驟。雖然該方法的識(shí)別率尚可接受,但是其魯棒性不高,且工作量巨大。此外,實(shí)際場(chǎng)景中目標(biāo)之間多有遮擋、背景復(fù)雜多樣以及拍攝角度不固定等因素都造成傳統(tǒng)方法識(shí)別困難甚至失效。本文旨在利用卷積神經(jīng)網(wǎng)絡(luò)(Convolutional Neural Networks,CNN)改善傳統(tǒng)行為識(shí)別方法中存在的這些問(wèn)題,在提高算法魯棒性的同時(shí)盡量提高識(shí)別的準(zhǔn)確率。針對(duì)背景減差法和幀間差分法在運(yùn)動(dòng)幅度不太大的情況下無(wú)法提取完整前景的缺點(diǎn),本文提出基于高斯差分(Difference of Gaussian,DoG)圖像的人體剪影提取算法。該方法利用兩張相鄰高斯尺度空間的圖像相減構(gòu)造包含人體輪廓信息的差分圖像,然后對(duì)其進(jìn)行二值強(qiáng)化、形態(tài)學(xué)處理等操作得到粗略的人體剪影圖像;第二步使用閾值對(duì)每行的粗略人體剪影區(qū)域進(jìn)行掃描檢測(cè),再經(jīng)閉運(yùn)算等操作后得到完整準(zhǔn)確的人體剪影圖像。為融合圖像序列的時(shí)域信息,本文累加周期內(nèi)的人體剪影圖像,生成二維特征圖,并將其送入到CNN中進(jìn)行訓(xùn)練識(shí)別。最終,經(jīng)過(guò)網(wǎng)絡(luò)調(diào)參和五折交叉驗(yàn)證等實(shí)驗(yàn)后在KTH公共數(shù)據(jù)集上得到85.3%的平均準(zhǔn)確率,證明該識(shí)別框架具有一定的可行性。為了更好地處理視頻數(shù)據(jù),學(xué)者們將卷積神經(jīng)網(wǎng)絡(luò)擴(kuò)展到了三維。本文利用3D CNN進(jìn)行實(shí)驗(yàn),發(fā)現(xiàn)特征組合"光流圖-幀差圖-三幀幀差圖"可以取得最佳識(shí)別效果。經(jīng)過(guò)網(wǎng)絡(luò)調(diào)參和五折交叉驗(yàn)證等實(shí)驗(yàn)后在KTH公共數(shù)據(jù)集上得到92.0%的平均準(zhǔn)確率。其次,通過(guò)分析KTH數(shù)據(jù)集中各類(lèi)樣本數(shù)量的比例分布及其對(duì)應(yīng)的準(zhǔn)確率,本論文提出使用二次訓(xùn)練、過(guò)取樣策略和擴(kuò)展數(shù)據(jù)集這三種改進(jìn)方法來(lái)證明數(shù)據(jù)分布不均衡對(duì)實(shí)驗(yàn)結(jié)果確有影響,并以此提高識(shí)別率。最終,三種改進(jìn)方法分別達(dá)到93.5%、92.8%和94.7%的平均準(zhǔn)確率,為小樣本或不均衡數(shù)據(jù)集的分類(lèi)問(wèn)題提供解決辦法。此外,利用3DCNN進(jìn)行行為識(shí)別的方法在減少特征提取工作量的同時(shí)提高了算法的魯棒性,即改善了傳統(tǒng)識(shí)別方法中存在的問(wèn)題。
[Abstract]:In recent years, with the introduction of high-definition video equipment, artificial intelligence based on behavior recognition technology has been rapidly developed in the field of intelligent safe city, smart home and military security. Because of its wide application prospect and economic value, behavior analysis and recognition technology has become a hotspot in the field of computer vision. The traditional behavior recognition algorithms are usually divided into three steps: motion foreground detection, feature extraction and training recognition. Although the recognition rate of this method is acceptable, its robustness is not high and the workload is enormous. In addition, many factors such as occlusion between targets, complex background and uncertain shooting angle in the actual scene result in difficulty or even invalidation of traditional methods. This paper aims to improve these problems in traditional behavior recognition methods by using convolution neural network (Convolutional Neural Networks,CNN) to improve the robustness of the algorithm and improve the accuracy of recognition as much as possible. Aiming at the disadvantage that background subtraction and inter-frame differential can not extract the complete foreground without too much motion amplitude, this paper proposes a human body silhouette extraction algorithm based on Gao Si differential (Difference of Gaussian,DoG image. In this method, two subtraction images of adjacent Gao Si scale space are used to construct differential images containing human contour information, and then binary enhancement and morphological processing are performed to obtain rough human silhouette images. In the second step, the threshold is used to scan and detect the rough body silhouette area of each line, and then the complete and accurate human body silhouette image is obtained after the operations such as blocking operation. In order to fuse the temporal information of the image sequence, the human body silhouette image is accumulated in the period, and the two-dimensional feature map is generated, which is sent into the CNN for training and recognition. Finally, the average accuracy rate of 85.3% is obtained on the KTH common data set by the experiments of network parameter adjustment and 50% discount cross-validation, which proves the feasibility of the recognition framework. In order to better deal with video data, researchers extend the convolution neural network to 3 D. In this paper, 3D CNN is used to carry out experiments and it is found that the best recognition effect can be obtained by combining "optical flow graph, frame difference graph and three frame difference map". The average accuracy is 92.0% on the KTH common data set after the experiments of network parameter adjustment and 50% discount cross-validation. Secondly, by analyzing the proportional distribution of the number of samples in the KTH data set and the corresponding accuracy, this paper proposes the use of secondary training. The oversampling strategy and the extended data set are three improved methods to prove that the uneven distribution of the data has an effect on the experimental results, and thus to improve the recognition rate. Finally, the three improved methods reach the average accuracy of 93.5% and 94.7% respectively, which provide a solution to the classification problem of small sample or unbalanced data set. In addition, the method of behavior recognition using 3DCNN not only reduces the workload of feature extraction, but also improves the robustness of the algorithm, that is, it improves the problems existing in the traditional recognition methods.
【學(xué)位授予單位】：中國(guó)科學(xué)技術(shù)大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類(lèi)號(hào)】：TP391.41;TP183

【參考文獻(xiàn)】

相關(guān)期刊論文 前8條

1 劉琮;許維勝;吳啟迪;;時(shí)空域深度卷積神經(jīng)網(wǎng)絡(luò)及其在行為識(shí)別上的應(yīng)用[J];計(jì)算機(jī)科學(xué);2015年07期

2 黃凱奇;陳曉棠;康運(yùn)鋒;譚鐵牛;;智能視頻監(jiān)控技術(shù)綜述[J];計(jì)算機(jī)學(xué)報(bào);2015年06期

3 陶新民;郝思媛;張冬雪;徐鵬;;不均衡數(shù)據(jù)分類(lèi)算法的綜述[J];重慶郵電大學(xué)學(xué)報(bào)(自然科學(xué)版);2013年01期

4 雷慶;李紹滋;;動(dòng)作識(shí)別中局部時(shí)空特征的運(yùn)動(dòng)表示方法研究[J];計(jì)算機(jī)工程與應(yīng)用;2010年34期

5 卜濤濤;盧超;;圖像分割算法研究[J];電腦知識(shí)與技術(shù);2010年08期

6 田國(guó)會(huì);;家庭服務(wù)機(jī)器人研究前景廣闊[J];國(guó)際學(xué)術(shù)動(dòng)態(tài);2007年01期

7 何衛(wèi)華;李平;文玉梅;葉波;;復(fù)雜背景下基于圖像融合的運(yùn)動(dòng)目標(biāo)輪廓提取算法[J];計(jì)算機(jī)應(yīng)用;2006年01期

8 范瑩,郭成安;一種運(yùn)動(dòng)圖像的檢測(cè)與識(shí)別技術(shù)[J];大連理工大學(xué)學(xué)報(bào);2004年01期

相關(guān)博士學(xué)位論文 前2條

1 曹鵬;不均衡數(shù)據(jù)分類(lèi)方法的研究[D];東北大學(xué);2014年

2 何衛(wèi)華;人體行為識(shí)別關(guān)鍵技術(shù)研究[D];重慶大學(xué);2012年

相關(guān)碩士學(xué)位論文 前7條

1 張永剛;基于多特征融合的行為識(shí)別算法研究[D];內(nèi)蒙古大學(xué);2016年

2 陸霖霖;基于改進(jìn)ISA深度網(wǎng)絡(luò)的人體行為識(shí)別研究與實(shí)現(xiàn)[D];電子科技大學(xué);2016年

3 吳杰;基于卷積神經(jīng)網(wǎng)絡(luò)的行為識(shí)別研究[D];電子科技大學(xué);2015年

4 費(fèi)凡;智能視頻監(jiān)控中運(yùn)動(dòng)人體異常行為的自動(dòng)檢測(cè)與識(shí)別算法的研究與實(shí)現(xiàn)[D];南京郵電大學(xué);2014年

5 郭勁智;視頻圖像行人檢測(cè)方法研究[D];華南理工大學(xué);2012年

6 覃耀輝;視頻中的人體動(dòng)作行為識(shí)別研究[D];電子科技大學(xué);2011年

7 趙鳳娟;行人異常智能視頻監(jiān)控系統(tǒng)研究與實(shí)現(xiàn)[D];電子科技大學(xué);2011年

，

本文編號(hào)：2274394