發(fā)布時(shí)間：2018-03-03 13:30

  本文選題：人臉識(shí)別　切入點(diǎn)：數(shù)據(jù)清理　出處：《西南交通大學(xué)》2017年碩士論文　論文類型：學(xué)位論文

【摘要】：近年來(lái)隨著深度卷積神經(jīng)網(wǎng)絡(luò)(deep convolution neural networks,DCNN)的引入,人臉識(shí)別的準(zhǔn)確率得以跨越式提升,各類相關(guān)應(yīng)用如人臉識(shí)別考勤,考生身份驗(yàn)證,刷臉支付,人臉歸類查詢等已開(kāi)始逐步投入使用,效果顯著。然而現(xiàn)實(shí)場(chǎng)景中非限定條件的人臉識(shí)別技術(shù)卻存在較多難題,例如姿態(tài)、光照、遮擋等困難,識(shí)別精度也隨著數(shù)據(jù)規(guī)模的增加和識(shí)別難度的增大而快速下降。目前的解決方案是通過(guò)增大復(fù)雜場(chǎng)景下的訓(xùn)練數(shù)據(jù)庫(kù)規(guī)模來(lái)學(xué)習(xí)到盡可能多的場(chǎng)景。然而大多數(shù)大型的數(shù)據(jù)庫(kù)(百萬(wàn)級(jí))由私人公司持有且不公開(kāi),即使目前公開(kāi)的大型數(shù)據(jù)庫(kù),也由于標(biāo)注信息過(guò)少、準(zhǔn)確性得不到較好保證,會(huì)影響DCNN的訓(xùn)練。本文從確保訓(xùn)練數(shù)據(jù)庫(kù)的準(zhǔn)確性和解決部分姿態(tài)問(wèn)題出發(fā),主要做出了以下工作:本文首先深入了解了深度學(xué)習(xí)的部分理論知識(shí),總結(jié)了當(dāng)前主流的深度學(xué)習(xí)開(kāi)源框架和人臉識(shí)別開(kāi)源項(xiàng)目以及國(guó)內(nèi)當(dāng)前主流人臉識(shí)別相關(guān)公司和解決方案。通過(guò)檢測(cè)誤差、檢測(cè)效果和效率以及在大規(guī)模數(shù)據(jù)上的穩(wěn)定性的實(shí)驗(yàn)對(duì)比分析了當(dāng)前主流的人臉檢測(cè)算法的性能,并根據(jù)實(shí)際情況提出了檢測(cè)評(píng)價(jià)機(jī)制減少誤差,最終選擇針對(duì)大規(guī)模數(shù)據(jù)進(jìn)行處理綜合效果好的算法,提出了基于關(guān)鍵點(diǎn)映射的人臉圖像歸一化算法。針對(duì)大規(guī)模數(shù)據(jù)庫(kù)準(zhǔn)確性無(wú)法保證、存在噪聲等問(wèn)題,提出了基于多角度評(píng)價(jià)的數(shù)據(jù)清理方法,在同一個(gè)類別中每張圖像與其他圖像進(jìn)行相似度計(jì)算,并統(tǒng)計(jì)與該圖像不相似的圖像數(shù)量,超過(guò)一定的數(shù)量就對(duì)該圖像進(jìn)行清理。通過(guò)多方面的實(shí)驗(yàn)驗(yàn)證了清理數(shù)據(jù)方法的有效性。實(shí)驗(yàn)證明,清理后的數(shù)據(jù)庫(kù)訓(xùn)練模型在LFW數(shù)據(jù)集上的準(zhǔn)確率得到了提升,以較小規(guī)模的訓(xùn)練集取得了 99.17%的準(zhǔn)確率,在Youtube數(shù)據(jù)集取得了 93.54%的準(zhǔn)確率。為了解決非限定人臉識(shí)別中的多姿態(tài)問(wèn)題,使用基于三維人臉模型的圖像校正生成正面圖像,并提出了將正面合成圖像提取的特征與原始圖像進(jìn)行特征線性融合的方法來(lái)生成新的特征向量。合成正面圖像能提供原始圖像不具備的特征,也存在信息丟失,而原始圖像的特征具有較高的參考價(jià)值,因此融合二者特征的新特征具有更全面的特征。實(shí)驗(yàn)證明,新的特征向量能夠有效提高人臉識(shí)別率,將原有的LFW上50對(duì)錯(cuò)誤匹配對(duì)矯正了 15對(duì),在SWJTU-MF DB上也取得了顯著的效果。
[Abstract]:In recent years, with the introduction of deep convolution neural networks (DCNN), the accuracy of face recognition has been improved by leaps and bounds. Face classification and query have been gradually put into use, and the effect is remarkable. However, there are many difficult problems in the face recognition technology, such as pose, illumination, occlusion and so on. Recognition accuracy also decreases rapidly with the increase of data size and difficulty. The current solution is to learn as many scenarios as possible by increasing the size of the training database in complex scenarios. Large databases (million levels) are held by private companies and are not publicly available, Even if the large database is open at present, the accuracy can not be guaranteed well because of too little tagged information, which will affect the training of DCNN. This paper starts from ensuring the accuracy of the training database and solving some posture problems. The main work is as follows: first of all, this paper has a deep understanding of some of the theoretical knowledge of in-depth learning, Summarizes the current mainstream deep learning open source framework and face recognition open source projects, as well as domestic mainstream face recognition related companies and solutions. The performance of the current mainstream face detection algorithms is compared and analyzed in the experiments of detection effect, efficiency and stability on large scale data. According to the actual situation, the detection evaluation mechanism is proposed to reduce the error. Finally, an algorithm for processing large scale data is selected, and a face image normalization algorithm based on key point mapping is proposed. The accuracy of large-scale database can not be guaranteed, and there are some problems, such as noise, etc. A data cleaning method based on multi-angle evaluation is proposed, in which the similarity between each image and other images in the same category is calculated, and the number of images that are not similar to the image is counted. More than a certain number of images are cleaned. The validity of the data cleaning method is verified by experiments in many aspects. The experimental results show that the accuracy of the database training model after cleaning is improved on the LFW dataset. In order to solve the problem of multi-pose in unqualified face recognition, the accuracy rate of 99.17% is obtained with the smaller training set and 93.54% with the Youtube dataset. In order to solve the problem of multi-pose in unqualified face recognition, the image correction based on 3D face model is used to generate the frontal image. A new feature vector is generated by linear fusion of features extracted from frontal composite image and original image. The synthesized frontal image can provide features that the original image does not possess, and there is also information loss. However, the features of the original image are of high reference value, so the new features with the fusion of the two features have more comprehensive features. Experiments show that the new feature vector can effectively improve the face recognition rate. The 50 pairs of error matching pairs on the original LFW were corrected by 15 pairs, and a remarkable effect was obtained on SWJTU-MF DB.
【學(xué)位授予單位】：西南交通大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：TP391.41

【參考文獻(xiàn)】

相關(guān)期刊論文 前3條

1 郭麗麗;丁世飛;;深度學(xué)習(xí)研究進(jìn)展[J];計(jì)算機(jī)科學(xué);2015年05期

2 余凱;賈磊;陳雨強(qiáng);徐偉;;深度學(xué)習(xí)的昨天、今天和明天[J];計(jì)算機(jī)研究與發(fā)展;2013年09期

3 葉曉蕓;秦鑒;;論淺層學(xué)習(xí)與深度學(xué)習(xí)[J];軟件導(dǎo)刊;2006年02期

，

本文編號(hào)：1561231