【摘要】：人臉檢測(cè)技術(shù)是模式識(shí)別領(lǐng)域的重要研究課題之一。在實(shí)際應(yīng)用中,采集到的人臉圖像往往會(huì)受到周圍環(huán)境的影響,造成人臉檢測(cè)中的姿態(tài)變化、遮擋和復(fù)雜背景等問(wèn)題,導(dǎo)致人臉檢測(cè)的準(zhǔn)確性和魯棒性有時(shí)不能滿足實(shí)際應(yīng)用的需求。結(jié)合深度學(xué)習(xí)理論,本文提出了一種基于深度學(xué)習(xí)混合模型的人臉檢測(cè)算法。通過(guò)建立深度學(xué)習(xí)混合模型,利用各特征之間的強(qiáng)相互關(guān)系學(xué)習(xí)人臉局部特征及位置,期望以此減少部分遮擋和多姿態(tài)對(duì)人臉檢測(cè)造成的影響。本文主要研究?jī)?nèi)容如下:1.首先結(jié)合深度學(xué)習(xí)理論,從特征提取、訓(xùn)練和收斂時(shí)間等角度對(duì)深度學(xué)習(xí)的三種結(jié)構(gòu)及其典型模型進(jìn)行了分析、研究。其次,從分類誤差和收斂性兩種角度對(duì)深度置信網(wǎng)絡(luò)和卷積神經(jīng)網(wǎng)絡(luò)兩種方法進(jìn)行了仿真對(duì)比實(shí)驗(yàn)。實(shí)驗(yàn)結(jié)果表明,使用卷積神經(jīng)網(wǎng)絡(luò)方法的分類誤差百分比在整體上要低于使用深度置信網(wǎng)絡(luò)方法的分類誤差百分比,但在收斂速度上,深度置信網(wǎng)絡(luò)的收斂速度要優(yōu)于卷積神經(jīng)網(wǎng)絡(luò)的收斂速度。因此,根據(jù)兩種典型模型的優(yōu)點(diǎn)與不足,本文構(gòu)建了一種卷積池化受限玻爾茲曼機(jī)模型單元,將卷積神經(jīng)網(wǎng)絡(luò)中的卷積層和池化層加入到受限玻爾茲曼機(jī)中的隱藏層中,構(gòu)建深度學(xué)習(xí)混合模型的基本單元。改進(jìn)后的網(wǎng)絡(luò)無(wú)需預(yù)處理就可以直接輸入原始圖像,其結(jié)構(gòu)更符合圖像輸入的拓?fù)浣Y(jié)構(gòu),更適合對(duì)圖像的訓(xùn)練學(xué)習(xí)。2.針對(duì)單一深度模型在解決人臉檢測(cè)部分遮擋時(shí)出現(xiàn)學(xué)習(xí)效率低、人臉檢測(cè)誤檢率高等問(wèn)題,提出了一種基于深度學(xué)習(xí)的混合模型——卷積池化深度置信網(wǎng)絡(luò)(Convolutional pooling deep belief network,CPDBN)算法來(lái)解決人臉檢測(cè)問(wèn)題。首先將之前構(gòu)造的卷積池化受限玻爾茲曼機(jī)作為深度模型的基本單元,然后建立多層基本單元結(jié)構(gòu),并利用深度模型深層結(jié)構(gòu)之間的相關(guān)性,學(xué)習(xí)各特征的位置和特征之間的相關(guān)關(guān)系。在檢測(cè)出現(xiàn)遮擋情況時(shí),根據(jù)學(xué)習(xí)檢測(cè)到的人臉局部特征,預(yù)測(cè)推斷隱藏的特征位置,由完整的人臉特征進(jìn)行人臉檢測(cè)。實(shí)驗(yàn)結(jié)果表明,本文算法加快了收斂速度,提高了人臉部分遮擋情況下的人臉檢測(cè)精度,而且對(duì)于姿態(tài)變化具有一定的魯棒性。
[Abstract]:Face detection is one of the important research topics in the field of pattern recognition. In practical applications, the collected face images are often affected by the surrounding environment, resulting in the changes of face pose, occlusion and complex background, and so on. As a result, the accuracy and robustness of face detection sometimes can not meet the needs of practical applications. Based on the theory of depth learning, a face detection algorithm based on the hybrid model of depth learning is proposed in this paper. In order to reduce the influence of partial occlusion and multi-pose on face detection, a hybrid model of depth learning is established to study the local features and location of human face by using the strong interrelation between each feature. The main contents of this paper are as follows: 1. Based on the theory of depth learning, three kinds of structures and their typical models of deep learning are analyzed and studied from the aspects of feature extraction, training and convergence time. Secondly, two methods of depth confidence network and convolutional neural network are simulated and compared with each other in terms of classification error and convergence. The experimental results show that the percentage of classification error using convolution neural network method is lower than that of using depth confidence network method, but the convergence rate is higher. The convergence speed of depth confidence network is better than that of convolution neural network. Therefore, according to the advantages and disadvantages of the two typical models, a convolution pool constrained Boltzmann machine model unit is constructed in this paper. The convolution layer and the pool layer in the convolution neural network are added to the hidden layer in the constrained Boltzmann machine. The basic unit of the hybrid model of deep learning is constructed. The improved network can input the original image directly without preprocessing, its structure is more consistent with the topological structure of image input, and it is more suitable for image training and learning. The single depth model can solve the problems of low learning efficiency and high false detection rate in face detection. In this paper, a hybrid model based on depth learning, convolution pool depth confidence network (Convolutional pooling deep belief network), is proposed to solve the problem of face detection. Firstly, the previously constructed convolution pool constrained Boltzmann machine is regarded as the basic unit of the depth model, and then the multilayer basic unit structure is established, and the correlation between the deep structure of the depth model is used. Learn the correlation between the location of each feature and the feature. When the occlusion is detected, the location of the hidden feature is predicted according to the local features detected by learning, and the face is detected by the complete face feature. Experimental results show that the proposed algorithm can accelerate the convergence speed, improve the accuracy of face detection under partial occlusion, and is robust to the change of pose.
【學(xué)位授予單位】：蘭州理工大學(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)】

中國(guó)期刊全文數(shù)據(jù)庫(kù) 前5條

1 張春霞;姬楠楠;王冠偉;;受限波爾茲曼機(jī)[J];工程數(shù)學(xué)學(xué)報(bào);2015年02期

2 顧偉;劉文杰;朱忠浩;許凱;;一種基于膚色模型和模板匹配的人臉檢測(cè)算法[J];微型電腦應(yīng)用;2014年07期

3 孫志軍;薛磊;許陽(yáng)明;王正;;深度學(xué)習(xí)研究綜述[J];計(jì)算機(jī)應(yīng)用研究;2012年08期

4 張學(xué)工;關(guān)于統(tǒng)計(jì)學(xué)習(xí)理論與支持向量機(jī)[J];自動(dòng)化學(xué)報(bào);2000年01期

5 盧春雨,張長(zhǎng)水,聞芳,閻平凡;基于區(qū)域特征的快速人臉檢測(cè)法[J];清華大學(xué)學(xué)報(bào)(自然科學(xué)版);1999年01期

中國(guó)碩士學(xué)位論文全文數(shù)據(jù)庫(kù) 前9條

1 許蕓玉;非限制條件下的多姿態(tài)人臉檢測(cè)算法研究[D];蘭州理工大學(xué);2016年

2 呂澤江;基于POCS算法的人臉檢測(cè)研究[D];吉林大學(xué);2016年

3 奉俊鵬;基于非下采樣Contourlet梯度方向直方圖的人臉識(shí)別方法研究[D];湘潭大學(xué);2015年

4 趙志國(guó);基于深度學(xué)習(xí)的低分辯率多恣態(tài)人臉識(shí)別[D];大連理工大學(xué);2015年

5 劉智;人臉檢測(cè)算法的并行化研究與實(shí)現(xiàn)[D];中南大學(xué);2014年

6 祁佳;視頻中人臉檢索與事件檢測(cè)技術(shù)研究[D];西安電子科技大學(xué);2013年

7 張志偉;基于人臉識(shí)別的媒資視頻檢索技術(shù)的研究與實(shí)踐[D];北京郵電大學(xué);2013年

8 董立新;基于先驗(yàn)知識(shí)的人臉檢測(cè)算法研究[D];大連理工大學(xué);2010年

9 龍敏;基于多示例學(xué)習(xí)的Adaboost算法及其在人臉檢測(cè)中的應(yīng)用[D];上海交通大學(xué);2007年

，

本文編號(hào)：2142039