文档库 最新最全的文档下载
当前位置:文档库 › 人脸识别外文文献

人脸识别外文文献

人脸识别外文文献
人脸识别外文文献

Method of Face Recognition Based on Red-Black

Wavelet Transform and PCA

Yuqing He, Huan He, and Hongying Yang

Department of Opto-Electronic Engineering,

Beijing Institute of Technology, Beijing, P.R. China, 100081

20701170@https://www.wendangku.net/doc/b71276563.html,

Abstract. With the development of the man-machine interface and the recogni-tion technology, face recognition has became one of the most important research aspects in the biological features recognition domain. Nowadays, PCA(Principal Components Analysis) has applied in recognition based on many face database and achieved good results. However, PCA has its limitations: the large volume of computing and the low distinction ability. In view of these limitations, this paper puts forward a face recognition method based on red-black wavelet transform and PCA. The improved histogram equalization is used to realize image pre-processing in order to compensate the illumination. Then, appling the red-black wavelet sub-band which contains the information of the original image to extract the feature and do matching. Comparing with the traditional methods, this one has better recognition rate and can reduce the computational complexity.

Keywords: Red-black wavelet transform, PCA, Face recognition, Improved histogram equalization.

1Introduction

Because the traditional status recognition (ID card, password, etc) has some defects, the recognition technology based on biological features has become the focus of the re-search. Compared with the other biological features (such as fingerprints, DNA, palm prints, etc) recognition technology, people identify with the people around mostly using the biological characteristics of human face. Face is the most universal mode in human vision. The visual information reflected by human face in the exchange and contact of people has an important role and significance. Therefore, face recognition is the easiest way to be accepted in the identification field and becomes one of most potential iden-tification authentication methods. Face recognition technology has the characteristics of convenient access, rich information. It has wide range of applications such as iden-tification, driver's license and passport check, banking and customs control system, and other fields[1].

The main methods of face recognition technology can be summed up to three kinds: based on geometric features, template and model separately. The PCA face recognition method based on K-L transform has been concerned since the 1990s. It is simple, fast. and easy to use. It can reflect the person face's characteristic on the whole. Therefore, applying PCA method in the face recognition is unceasingly improving.

D.-S. Huang et al. (Eds.): ICIC 2008, LNCS 5226, pp. 561–568, 2008. ? Springer-Verlag Berlin Heidelberg 2008

This paper puts forward a method of face recognition based on Red-Black wavelet transform and PCA. Firstly, using the improved image histogram equalization[2] to do image preprocessing, eliminating the impact of the differences in light intensity. Sec-ondly, using the Red-Black wavelet transform to withdraw the blue sub-band of the relative stable face image to obscure the impacts of expressions and postures. Then, using PCA to withdraw the feature component and do recognition. Comparing with the traditional PCA methods, this one can obviously reduce computational complexity and increase the recognition rate and anti-noise performance. The experimental results show that this method mentioned in this paper is more accurate and effective.

2Red-Black Wavelet Transform

Lifting wavelet transform is an effective wavelet transform which developed rapidly these years. It discards the complex mathematical concepts and the telescopic and translation of the Fourier transform analysis in the classical wavelet transform. It de-velops from the thought of the classical wavelet transform multi-resolution analysis. Red-black wavelet transform[3-4] is a two-dimensional lifting wavelet transform[5-6], it contains horizontal/vertical lifing and diagonal lifting. The specific principles are as bellow.

2.1Horizontal /Vertical Lifting

As Fig.1 shows, horizontal /vertical lifting is divided into three steps:

1. Decomposition: The original image by horizontal and vertical

direction is divided into red and black block in

a cross-block way.

2. Prediction: Carry on the prediction using horizontal and the

vertical direction four neighborhood's red

blocks to obtain a black block predicted value.

Then, using the difference of the black block actual value and the predicted value to

substitute the black block actual value. Its

result obtains the original image wavelet

coefficient. As Fig.1(b) shows:

[]

←--+-++++

f i j f i j f i j f i j f i j f i j

(,)(,)(1,)(,1)(,1)(1,)/4

i j

≠(1)

(mod2mod2)

3. Revision:

Using the horizontal and vertical direction four neighborhood's black block's wavelet coefficient to revise the red block actual value to obtain the approximate signal. As Fig.1(c) shows:

[]

←+-+-++++

f i j f i j f i j f i j f i j f i j

(,)(,)(1,)(,1)(,1)(1,)/8

=(2)

i j

(mod2mod2) In this way, the red block corresponds to the approximating information of the image, and the black block corresponds to the details of the image.

2.2Diagonal Lifting

On the basis of horizontal /vertical lifting, we do the diagonal lifting. As Fig.2 shows, it is also divided into three steps:

Fig.2.Diagonal lifting

1.Decomposition: After horizontal /vertical lifting, dividing the

obtained red block into the blue block and the yellow block in the diagonal cross way.

2. Prediction: Using four opposite angle neighborhood's blue block to predict a data in order to obtain the yellow block predicted value. Then the difference of the yellow block actual value and the predicted value substitutes the yellow block actual value. Its result obtains the original image wavelet coefficient of the diagonal direction. As Fig.2(b) shows:

[]←---+-+++-+++ (,)(,)(1,1)(1,1)(1,1)(1,1)/4

f i j f i j f i j f i j f i j f i j

==(3)

(mod21,mod21)

i j

3. Revision: Using four opposite angle neighborhood yellow block wavelet co-efficient to revise the blue block actual value in order to obtain the approximate signal. As Fig.2(c) shows:

[]←---+-+++-+++ (,)(,)(1,1)(1,1)(1,1)(1,1)/8

f i j f i j f i j f i j f i j f i j

==(4)

(mod20,mod20)

i j

After the second lifting, the red-black wavelet transform is realized.

According to the Equations, it can analyze some corresponding relations between the red-black wavelet transform and the classical wavelet transform: namely, the blue block is equal to the sub-band LL of the classical tensor product wavelets, the yellow block is equal to sub-band HH and the black block is equal to sub-band HL and LH. Experimental results show that it discards the complex mathematical concepts and equations. The relativity of image can mostly be eliminated and the sparser represen-tation of image can be obtained by the Red-Black wavelet transform.

The image after Red-Black wavelet transform is showed in the Fig.3(b), on the left corner is the blue sub-band block image which is the approximate image of original image.

Fig.3.The result of red-black wavelet transform

3 Feature Extraction Based on PCA[7]

PCA is a method which analyses data in statistical way. This method discovers group of vectors in the data space. Using these vectors to express the data variance as far as possible. Putting the data from the P-dimensional space down to M-dimensional space ( P>>M). PCA use K-L transform to obtain the minimum-dimensional image recogni-tion space of the approximating image space. It views the face image as a high-dimensional vector. The high-dimensional vector is composed of each pixel. Then the high-dimensional information space maps the low-dimensional characteristic subspace by K-L transform. It obtains a group of orthogonal bases through high-dimensional face image space K-L transform. The partial retention of orthogonal bases creates the low-dimensional subspace. The orthogonal bases reserved is called “Principle component”. Since the image corresponding to the orthogonal bas es just like face, so this is also called “Eigenfaces” method. The arithmetic of feature extraction are specified as follows:

For a face image of m × n, connecting its each row will constitute a row vector which has D= m × n dimensions. The D is the face image dimensions. Supposing M is the number of training samples, x j is the face image vector which is derived from the jth picture, so the covariance matrix of the whole samples is:

1()()M

T T j j j s x u x u ==--∑ (5)

And the μ is the average image vector of the training samples:

11M j j u x

M ==∑ (6)

Ordering []12,,,,M A x u x u x u =--- so T T S AA =and its demision is D ?D.(7)

According to the principle of K-L transform, the coordinate we achieved is com-posed of eigenvector corresponding to nonzero eigenvalue

of matrix Computing out the eigenvalue and Orthogonal normalized vector of matrix D×D directly is diffi -cult. So according to the SVD principle, it can figure out the eigenvalue and eigen-vector of

matrix through getting the eigenvalue and eigenvector of matrix (1,2,,)i i r λ= is r nonzero eigenvalue of

matrix i v :is the

eigenvector corre-sponding to i λ so the orthogonal normalized

i λeigenvector of matrix is as bel-low:

This is the eigenvector of Arranging its eigenvalues according to the size: 12i λλλ≥≥≥ ,its corresponding eigenvector is i μ. In this

way, each face image can project on the sub-space composed of, 1μ2μ…r μ. In order to reduce the dimension, it can select the former d eigenvectors as sub-space. It can select d biggest eigenvectors according to the energy proportion which the eigenvalue occupies:

Usually ordering α=90%~99%.

As a result , the image corresponding to these eigenvectors are similar to the human face, it is also called “Eigenfaces”.So the method which uses PCA transform is called“Eigenfaces”method. Owing to the Drop-dimensional space composed of “Ei -genfaces”, each image can project on it and get a group of coordinate coefficients which shows the location of the sub-space of this image, so it can be used as the bases for face recognition. Therefore, it can use the easiest Nearest Neighbor Classifier[8] to classify the faces.

4 Experiments and Results

This part mainly verifies the feasibility and superiority of the algorithm through the comparison of the experimental data.

4.1 Experimental Conditions and Parameters

Selecting Images After Delamination. After Red-Black wavelet transform, we only select the data of the blue block, because the part of the blue block represents the ap-proximating face image and it is not sensitive to the expression and illumination and even filtrates the image noise.

The Influence of The Blue Block Energy Caused By Decomposition Layers. In the experiment data[9] of the Red-Black transform, we can find that it does not have the energy centralized characteristic under the situation of multi-layer decomposition. As the Table 1 shows that different layer decompositions obtain different energies. Test image is the international

standard image Lena512, its energy is 4.63431e+009 and entropy is 7.4455.

The original image energy wastage is due to the black and yellow block transform.

According to the former results and the size of the face image (112×92), one layer decomposition can be done to achieve satisfactory results. The blue block sub-band not only has no incentive to expression and gestures, but also retains the difference of different faces. At the same time it reduces the image vector dimensions and the com-plexity of the algorithm. If the size of the original image is bigger and the resolution is higher, the multi-layer decompositions can be considered.

Database Selection. We choose the public ORL database to do some related ex-periments. This database contains 40 different people’s images which are captured in different periods and situation. It has 10 pictures per person and 400 pictures in all. The background is black. Each picture has 256 grayscales and the size is 112×92. The face images in the database have the different facial expressions and the different facial detail changes. The facial postures also have the changes. At present, this is the most extensive face database.

4.2 Experiment Processes and Results

The method of face recognition based on Red-Black wavelet transform and PCA shows as bellow: Firstly, using the improved image histogram equalization to do pretreatment,

eliminating the impact of the differences in light intensity. Secondly, using the Red-black wavelet transform to withdraw the blue block sub-band of the relative stable person face image achieved the effects of obscuring impacts of expressions and pos-tures. Then, using PCA to withdraw the feature component and do recognition. We adopt 40 person and 5 pictures per person when training, so there are 200 pictures as the training samples all together. Then carrying on recognition to the other 200 pictures under the conditions of whether there are illumination and Red-Black wavelet

trans-form or not.

Image Preprocessing. First using the illumination compensation on the original face images (Fig.5) in the database, namely doing gray adjustment and normalization, the images (Fig.6) after transform are obviously clearer than the former ones and helpful for analysis and recognition. Put the one layer Red-Black wavelet transform on the com-pensation images, then withdraw the images of the blue block sub-band which are the low-dimension approximating images of the original images (Fig.7). Appling Red-Black wavelet transform on the images plays an important role in obscuring the impacts of face expressions and postures and achieved good effects of reducing dimensions.

Feature Extraction and Matcing Results.After the image preprocessing, we adopt PCA to extract features and recognize. This paper analyses the results of three models which separately are PCA combined with Red-Black wavelet transform, illumination compensation, Red-Black wavelet transform and illumination compensation. The recognition rates, training and recognition time of different models are showed in the Table 2. We can see that withdrawing the blue block sub-band can obviously reduces the dimensions of the image vector and computation. The reducing training time shows that the low resolution subgraph reduces the computational complexity of the PCA through the Red-Black wavelet transform. Since illumination have great influence on feature extraction and recognition based on PCA, so the recognition effects can be enhanced by illumination compensation. Therefore, the combination of Red-Black wavelet transform, illumination compensation and PCA can achieve more satisfactory system performance. Comparing with the traditional method using wavelet transform and PCA , the recognition rate is enhanced obviously.

5 Conclusion

The Red-Black wavelet transform divides the rectangular grid digital image into red and black blocks and uses the two-dimensional lifting form to construct sub-band. It is an effective way to wipe off the image relativity and gets the more sparser image. PCA extracts eigenvector on the basis of the whole face grayscale relativity. The eigenvector can retain the main classified information in the original image space and rebuild the original image at the lowest MSE. The method of extracting

face features through combining the Red-Black wavelet transform, illumination compensation and PCA can effectively reduce computational complexity at the guarantee of certain recognition rate. Experimental results prove that this scheme is effective in the face recognition.

Acknowledgement.This project is supported by National Science Foundation of China (No. 60572058) and Excellent Young Scholars Research Fund of Beijing Insti-tute of Technology (No. 2006Y0104).

References

1. Chellappa, R., Wilson, C., Sirohe, S.: Human and Machine Recognition of Faces: A Survey. Proceedings of the IEEE 83(5), 705–740 (1995)

2. Yao, R., Huang, J., Wu, X.Q.: An Improved Image Enhancement Algorithm of Histogram Equalization. Journal of The China Railway Society 19(6), 78–81 (1997)

3. Uytterhoeven, G., Bultheel, A.: The Red-Black Wavelet Transform. Katholieke Universiteit Leuven, Belgium (1977)

4. Uytterhoeven, G., Bultheel, A.: The Red-Black Wavelet Transform and the Lifting Scheme. Katholieke Universiteit Leuven, Belgium (2000)

5. Sweldens, W.: The lifting scheme: a Custom-Design Construction of Biorthogonal Wavelets. Applied and Computational Harmonic Analysis 3(2), 186–200 (1996)

6. Kovacvic, J., Sweldens, W.: Wavelet Families of Increasing order in Arbitrary Dimension. IEEE Transactions on Image Processing 9(3), 280–496 (2000)

7. Turk, M., Pentland, A.: Eigenfaces for Recognition. Cognitive Neuroscience 3(1), 71–86 (1991)

8. Stan, Z., Li, J., Lu: Face Recognition Using the Nearest Feeature Line Method. IEEE Trans. on Neural Network 10(2), 439–443 (1999)

9. Wang, H.: Improvement on Red-Black Wavelet Transform. Journal of zaozhuang univer-sity 24(5), 61–63 (2007)

NMF综述报告

人脸识别的非负矩阵分解(NMF)方法文献综述 摘要:人类对整体的感知是基于对部分的感知,NMF(非负矩阵分解,Non-negative matrix factorization)的思想正是源于此。通过对矩阵分解因子加入了非负性约束,使得对高维非负原始数据矩阵的分解结果不存在负值,且具有一定的稀疏性,因而得到了相对低维、纯加性、拥有一定稀疏特性的分解结果。与PCA(主成分分析,principal components analysis)等传统人脸识别方法相比,NMF的基图像就是人脸的各个局部特征,并且通过对经典算法的一系列优化,改进的NMF算法的识别率和鲁棒性较传统方法有着显著优势。此外,NMF在机器学习、语义理解等领域也有着重要应用。 关键词:非负矩阵分解(NMF)稀疏性改进的NMF 语义理解 一、引言 在实际中的许多数据都具有非负性,而现实中对数据的处理又要求数据的低秩性经典的数据处理方法一般不能够确保非负性的要求,如何找到一个非负的低秩矩阵来近似原数据矩阵成为一个关键问题。在这样的背景下,NMF方法应运而生。 NMF方法思想最早可以追溯到由Paatero和Tapper在1994年提出的正矩阵分解(Positive Matrix Factorization,PMF)[1];此后1999年,Lee和Seung提出了一个以广义KL散度为优化目标函数的基本NMF模型算法,并将其应用于人脸图像表示[2];2001年,Lee和Seung通过对基本NMF算法进行深入研究,又提出了两个经典的NMF算法,即基于欧氏距离测度的乘性迭代算法和基于广义KL散度的乘性迭代算法,并给出了收敛性证明[3],这两种算法称为NMF方法的基准算法,广泛应用于各个领域。 但是在实际应用中,由于经典的基准NMF算法存在收敛速度较慢,未利用统计特征,对光线、遮挡等敏感,以及无法进行增量学习等问题,各种改进的NMF算法被提出。其中包括Lin提出的基于投影梯度(Projected Gradient,PG)的NMF方法[3],该方法有着很高的分解精度;Berry提出的基于投影非负最小二乘(Projected Non-negative Least Square,PNLS)的NMF方法[5],通过这种方法得到的基矩阵的稀疏性、正交性叫基准NMF方法都更好;此外还有牛顿类方法[6]和基于有效集[7]的NMF方法等。 二、NMF的基准算法 1.NMF模型 给定一个非负矩阵(即),和一个正整数,求未知非负矩阵和,使得 用表示逼近误差矩阵。可以用下图表示该过程:

人工智能专业外文翻译-机器人

译文资料: 机器人 首先我介绍一下机器人产生的背景,机器人技术的发展,它应该说是一个科学技术发展共同的一个综合性的结果,同时,为社会经济发展产生了一个重大影响的一门科学技术,它的发展归功于在第二次世界大战中各国加强了经济的投入,就加强了本国的经济的发展。另一方面它也是生产力发展的需求的必然结果,也是人类自身发展的必然结果,那么随着人类的发展,人们在不断探讨自然过程中,在认识和改造自然过程中,需要能够解放人的一种奴隶。那么这种奴隶就是代替人们去能够从事复杂和繁重的体力劳动,实现人们对不可达世界的认识和改造,这也是人们在科技发展过程中的一个客观需要。 机器人有三个发展阶段,那么也就是说,我们习惯于把机器人分成三类,一种是第一代机器人,那么也叫示教再现型机器人,它是通过一个计算机,来控制一个多自由度的一个机械,通过示教存储程序和信息,工作时把信息读取出来,然后发出指令,这样的话机器人可以重复的根据人当时示教的结果,再现出这种动作,比方说汽车的点焊机器人,它只要把这个点焊的过程示教完以后,它总是重复这样一种工作,它对于外界的环境没有感知,这个力操作力的大小,这个工件存在不存在,焊的好与坏,它并不知道,那么实际上这种从第一代机器人,也就存在它这种缺陷,因此,在20世纪70年代后期,人们开始研究第二代机器人,叫带感觉的机器人,这种带感觉的机器人是类似人在某种功能的感觉,比如说力觉、触觉、滑觉、视觉、听觉和人进行相类比,有了各种各样的感觉,比方说在机器人抓一个物体的时候,它实际上力的大小能感觉出来,它能够通过视觉,能够去感受和识别它的形状、大小、颜色。抓一个鸡蛋,它能通过一个触觉,知道它的力的大小和滑动的情况。第三代机器人,也是我们机器人学中一个理想的所追求的最高级的阶段,叫智能机器人,那么只要告诉它做什么,不用告诉它怎么去做,它就能完成运动,感知思维和人机通讯的这种功能和机能,那么这个目前的发展还是相对的只是在局部有这种智能的概念和含义,但真正完整意义的这种智能机器人实际上并没有存在,而只是随着我们不断的科学技术的发展,智能的概念越来越丰富,它内涵越来越宽。 下面我简单介绍一下我国机器人发展的基本概况。由于我们国家存在很多其

指纹识别系统(文献综述)

指纹识别方法的综述 摘 要: 对在指纹的预处理和特征提取、指纹分类、指纹的匹配过程中的方向图、滤波器、神经网络等关 键性原理和技术做了详细的说明,并对在各个过程中用到的方法做了进一步的比较,讨论了各种方法的优越性。 0 引 言 自动指纹识别是上世纪六十年代兴起的,利用计算机取代人工来进行指纹识别的一种方法。近年 来,随着计算机技术的飞速发展,低价位指纹采集仪的出现以及高可靠算法的实现,更使得自动指纹识 别技术越来越多地进入到人们的生活和工作中,自动指纹识别系统的研究和开发正在成为国内外学术 界和商业界的热点。相对于其他生物特征鉴别技术例如语音识别及虹膜识别,指纹识别具有许多独到 的优点,更重要的是它具有很高的实用性和可行性,已经被认为是一种理想的身份认证技术,有着十分 广泛的应用前景,是将来生物特征识别技术的主流。 1 指纹取像 图 1 是一个自动指纹识别系统AFIS(Automated Fingerprint Identification System) 的简单流程。 → → → ↓ ↑ ———— 将一个人的指纹采集下来输入计算机进行处理是指纹自动识别的首要步骤。指纹图像的获取主要利用设备取像,方便实用,比较适合AFIS 。利用设备取像的主要方法又利用光学设备、晶体传感器和超声波来进行。光学取像设备是根据光的全反射原理来设计的。晶体传感器取像是根据谷线和脊线皮肤与传感器之间距离不同而产生的电容不同来设计的。超声波设备取像也是采用光波来取像,但由于超声波波长较短,抗干扰能力较强,所以成像的质量非常好。 2 图像的预处理与特征提取 无论采取哪种方法提取指纹,总会给指纹图像带来各种噪声。预处理的目的就是去除图像中的噪 音,把它变成一幅清晰的点线图,以便于提取正确的指纹特征。预处理是指纹自动识别过程的第一步, 它的好坏直接影响着指纹识别的效果。常用的预处理与特征提取( Image Preprocessing and Feature Ex 2 t raction) 方法的主要步骤包括方向图计算、图像滤波、二值化、细化、提取特征和后处理。当然这些步骤 可以根据系统和应用的具体情况再进行适当变化。文献[ 1 ]提出了基于脊线跟踪的方法能够指纹取像 图像预处理 特征提取 指纹识别 数据库管理

中英文文献翻译-加工中心数控技术

加工中心数控技术 出处:数控加工中心的分类以及各自特点 出版社:化学工业出版社; 第1版 (2009年3月16日) 作者:徐衡、段晓旭 加工中心是典型的集高技术于一体的机械加工设备,它的发展代表了一个国家设计制造的水平也大大提高了劳动生产率,降低了劳动成本,改善了工人的工作环境,降低了工人的劳动强度。本文经过对不同运动方案和各部件的设计方案的定性分析比较确定该教立式加工中心的进给传动方案为:采用固定床身,电主轴通过安装座安装在床身导轨的滑座上,床身导轨采用滚动导轨,可以实现Y 方向的进给运动。由X-Y双向精密数控工作台带动工件完成X,Y两个方向的进给运动;X,Y,Z三个方向的进给运动均滚珠丝杠,并由交流伺服电机驱动。导轨、滚珠丝杠有相应的润滑、防护等装置。 加工中心(英文缩写为CNC 全称为Computerized Numerical Control):是带有刀库和自动换刀装置的一种高度自动化的多功能数控机床。在中国香港,台湾及广东一代也有很多人叫它电脑锣。 工件在加工中心上经一次装夹后,数字控制系统能控制机床按不同工序,自动选择和更换刀具,自动改变机床主轴转速、进给量和刀具相对工件的运动轨迹及其他辅助机能,依次完成工件几个面上多工序的加工。并且有多种换刀或选刀功能,从而使生产效率大大提高。 加工中心数控机床是一种装有计算机数字控制系统的机床,数控系统能够处理加工程序,控制机床完成各种动作。与普通机床相比,数控机床能够完成平面曲线和空间曲面的加工,加工精度和生产效率都比较高,因而应用日益广泛。 数控机床的组成 一般来说,数控机床由机械部分、数字控制计算机、伺服系统、PC控制部分、液压气压传动系统、冷却润滑和排泄装置组成。数控机床是由程序控制的,零件的编程工作是数控机床加工的重要组成部分。伺服系统是数控机床的驱动部分,计算机输出的控制命令是通过伺服系统产生坐标移动的。普通的立式加工中心有三个伺服电机,分别驱动纵向工作台、横向工作台、主轴箱沿X向、Y向、Z向运动。X、Y、Z是互相垂直的坐标轴,因而当机床三坐标联动时可以加工空

人脸识别技术发展及应用分析解读

人脸识别技术发展及应用分析 人脸识别是基于人的脸部特征信息进行身份识别的一种生物识别技术。用摄像机采集人脸图像,并自动在图像中检测和跟踪人脸,进而对检测到的人脸进行脸部的一系列相关技术处理,包括人脸图像采集、人脸定位、人脸识别预处理、记忆存储和比对辨识,达到识别不同人身份的目的。 市场现状 人脸识别技术的研究始于20世纪60年代末期。 20世纪90年代后期以来,一些商业性的人脸识 别系统逐渐进入市场,但是,这些技术和系统离 实用化都有一定距离,性能和准确率也有待提高。 美国遭遇恐怖袭击后,这一技术引起了广泛关 注。作为非常容易隐蔽使用的识别技术,人脸识 别逐渐成为国际反恐和安全防范重要的手段之一。 近年来,人脸识别在中国市场,也经历着迅速的 发展,而且发展的脚步也越来越快。主要原因有以下两方面。 科技的进步 国际上,美国标准与技术研究院(NIST)举办的Face Recognition Vendor Test 2006,通过大规模的人脸数据测试表明,当今世界上人脸识别方法的识别 精度比2002年的FRVT2002至少提高了一个数量级(10倍),而对于高清晰,高质量人脸图像识别,机器的识别精度几乎达到100%。在我国,近年来科技界和社会各个方面都认识到人脸识别技术的重要性,国家政策对人脸识别技术研究给予了很大支持,使得我国人脸识别技术也得到了迅速的发展。 应用需求的增加 越来越趋向于高科技的犯罪手段使得人们对各种场合的安全机制要求也近乎 苛刻,各种应用需求不断涌现。人脸识别市场的快速发展一方面归功于生物识别需求的多元化,另一方面则是由于人脸识别技术的进步。从需求上来说,除了传统的考勤、门禁等应用外,视频监控环境下的身份识别正成为一种迫切的需求,

基于matlab人脸识别技术 开题报告

毕业设计(论文)开题报告 毕业设计(论文)课题情况,根据所查阅的文献资料,每人撰写2500字以上的文献综述,文后应列出所查阅的文献资料。 基于matlab人脸识别技术的实现 文献综述 一、MATLAB概述 MATLAB是矩阵实验室(Matrix Laboratory)的简称,是美国MathWorks公司出品的商业数学软件,用于算法开发、数据可视化、数据分析以及数值计算的高级技术计算语言和交互式环境,主要包括MATLAB和Simulink两大部分。MATLAB主要面对科学计算、可视化以及交互式程序设计的高科技计算环境。它将数值分析、矩阵计算、科学数据可视化以及非线性动态系统的建模和仿真等诸多强大功能集成在一个易于使用的视窗环境中,为科学研究、工程设计以及必须进行有效数值计算的众多科学领域提供了一种全面的解决方案,并在很大程度上摆脱了传统非交互式程序设计语言(如C、Fortran)的编辑模式,代表了当今国际科学计算软件的先进水平。应用于工程计算、控制设计、信号处理与通讯、图像处理、信号检测、金融建模设计与分析等领域。而在本文中主要用到的功能是图像处理功能。 二、BP神经网络概述 人工神经网络(Artificial Neural Net works,简写为ANNs)也简称为神经网络(NNs)或称作连接模型(Connectionist Model),它是一种模范动物神经网络行为特征,进行分布式并行信息处理的算法数学模型。这种网络依靠系统的复杂程度,通过调整内部大量节点之间相互连接的关系,从而达到处理信息的目的。 人工神经网络发展的主要历程有:20世纪50年代末,Rosenblatt提出的感知器模型和Widrow提出的自适应线性元件,出现了简单的线性分类器;1986年,Rumelhart和Mcllelland 提出了层网络“误差反向传播算法(BP)”,使有导师学习多层感知器网络(ML PN)模式分类器走向实用化,在此基础上又派生出若干前向网络,如径向基函数网络( RBFN)和函数链网络等;1982年,美国加州工学院的物理学家Hopfield提出的一种用于联想记忆和优化计算的反馈网络模型,由于引进了“能量函数” 的概念,使网络走向具体电路有了保证;20世纪70年代,Watanabe 提出了使用模式子空间的概念来设计不同类别对应的子空间,由不同类别聚类的子空间实现模式识别; Kohonen提出的自组织特征映射网络模型等都为神经网络模式识别理论提供了进一步的根据。 构成人工神经网络的三个基本要素是:神经元、络拓扑结构和网络的训练(学习)方法。神经元(节点)的作用是把若干输入加权求和,并对这种加权和进行非线性处理后输出。神经元的选择一般有以下特点:每个神经元都具有多个输入、个输出,具有闭值,采用非线性函数。 1、神经元

人脸识别发展趋势及应用领域分析

人脸识别发展趋势及应用领域分析 人脸识别技术是一种高精度、易于使用、稳定性高、难仿冒的生物识别技术,具有极其广阔的市场应用前景。在公安、国防、海关、交通、金融、社保、医疗及其他民用安全控制等行业和部门存在着广泛的需求。 一、人脸识别何以瞬间爆发? 其实对于整个生物识别领域来说,由于指纹识别应用时间早,价格低廉,而且使用便利,因此早早地便占据了国内的大部分市场,在顶峰时期,甚至可以达到90%左右。但是,又是什么力量,使人脸识别在短短五六年的时间里,就实现了如此迅猛的腾飞呢? (一)政策因素:抛开技术因素,国家政策的支持可以说是人脸识别崛起的重要因素之一。尤其随着近些年来,平安城市等技术的大力推广,我国安防市场的需求也随之迅速升温,各领域安保的等级也就随之实现明显的提升。甚至在部分地区的部分领域,人脸识别已经开始被列为使用过程中的强制标准。这种情况的出现,对于人脸识别的推广无疑是一种非常有力的推动。因此,虽然政策的角色只是一种诱导的作用,但是这种诱导对于人脸识别的爆发,却又是不可或缺的。 (二)社会需求:在我国,随着城镇化的进程加快,流动人口比例大大增加,面临的突发事件和异常事件越来越复杂。因此这也就给城市的安保工作,以及视频监控带来了更大的压力。传统依靠人工来“盯”的方式难免会因疲劳或精神不集中等原因影响监视效果,难以胜任对庞大数据库的分析与理解,从而影响事后查找证据,更难以满足时代的需求,于是,人们对于具有智能分析的视频监控应用的呼声越来越高。 二、人脸识别发展趋势 (一)与视频监控相结合:随着人脸识别技术的进一步发展,将人脸识别技术将实现与数字监控系统的进一步融合,将成为人脸识别技术的另外一大应用领域。据统计数据显示,仅在中国大陆,在未来三年内有望形成年销售额过百亿,并在未来十年内则有望形成年销售额过千亿的市场规模。(二)逐步取代指纹考勤:人脸识别考勤,通过对人脸一些独一无二的特征识别对验进行考

文献综述_人工智能

人工智能的形成及其发展现状分析 冯海东 (长江大学管理学院荆州434023) 摘要:人工智能的历史并不久远,故将从人工智能的出现、形成、发展现 状及前景几个方面对其进行分析,总结其发展过程中所出现的问题,以及发展现状中的不足之处,分析其今后的发展方向。 关键词:人工智能,发展过程,现状分析,前景。 一.引言 人工智能最早是在1936年被英国的科学家图灵提出,并不为多数人所认知。 当时,他编写了一个下象棋的程序,这就是最早期的人工智能的应用。也有著名的“图灵测试”,这也是最初判断是否是人工智能的方案,因此,图灵被尊称为“人工智能之父”。人工智能从产生到发展经历了一个起伏跌宕的过程,直到目前为止,人工智能的应用技术也不是很成熟,而且存在相当的缺陷。 通过搜集的资料,将详细的介绍人工智能这个领域的具体情况,剖析其面临的挑战和未来的前景。 二.人工智能的发展历程 1. 1956年前的孕育期 (1) 从公元前伟大的哲学家亚里斯多德(Aristotle)到16世纪英国哲学家培根(F. Bacon),他们提出的形式逻辑的三段论、归纳法以及“知识就是力量”的警句,都对人类思维过程的研究产生了重要影响。 (2)17世纪德国数学家莱布尼兹(G..Leibniz)提出了万能符号和推理计算思想,为数理逻辑的产生和发展奠定了基础,播下了现代机器思维设计思想的种子。而19世纪的英国逻辑学家布尔(G. Boole)创立的布尔代数,实现了用符号语言描述人类思维活动的基本推理法则。 (3) 20世纪30年代迅速发展的数学逻辑和关于计算的新思想,使人们在计算机出现之前,就建立了计算与智能关系的概念。被誉为人工智能之父的英国天才的数学家图灵(A. Tur-ing)在1936年提出了一种理想计算机的数学模型,即图灵机之后,1946年就由美国数学家莫克利(J. Mauchly)和埃柯特(J. Echert)研制出了世界上第一台数字计算机,它为人工智能的研究奠定了不可缺少的物质基础。1950年图灵又发表了“计算机与智能”的论文,提出了著名的“图灵测试”,形象地指出什么是人工智能以及机器具有智能的标准,对人工智能的发展产生了极其深远的影响。 (4) 1934年美国神经生理学家麦克洛奇(W. McCulloch) 和匹兹(W. Pitts )建立了第一个神经网络模型,为以后的人工神经网络研究奠定了基础。 2. 1956年至1969年的诞生发育期 (1)1956年夏季,麻省理工学院(MIT)的麦卡锡(J.McCarthy)、明斯基(M. Minshy)、塞尔夫里奇(O. Selfridge)与索罗门夫(R. Solomonff)、 IBM的洛

虹膜识别外文翻译文献

虹膜识别外文翻译文献 虹膜识别外文翻译文献 (文档含中英文对照即英文原文和中文翻译) 外文: The first chapter 1.1 The research background of iris recognition Biometrics is a technology for personal identification using physiological characteristics and behavior characteristics inherent in the human body. Can be used for the biological characteristics of biological recognition, fingerprint, hand type face, iris, retina, pulse, ear etc.. Behavior has the following characteristics: signature, voice, gait, etc.. Based on these characteristics, it has been the development of hand shape recognition, fingerprint recognition, facial recognition, iris recognition, signature recognition and other biometric technology, many techniques have been formed and mature to application of. Biological recognition technology in a , has a long history, the ancient Egyptians through identification of each part of the body size measure to carry out identity may be the earliest human based on the earliest history of biometrics. But the modern biological recognition technology began in twentieth Century 70 time metaphase, as biometric devices early is relatively expensive, so only a higher security level atomic test, production base.due to declining cost of microprocessor and various electronic components, precision gradually improve, control device of a biological recognition technology has been gradually applied to commerce authorized, such as access control, attendance management, management system, safety certification field etc.. All biometric technology, iris recognition is currently used as a convenient and accurate.

机械类数控车床外文翻译外文文献英文文献车床.doc

Lathes Lathes are machine tools designed primarily to do turning, facing and boring, Very little turning is done on other types of machine tools, and none can do it with equal facility. Because lathes also can do drilling and reaming, their versatility permits several operations to be done with a single setup of the work piece. Consequently, more lathes of various types are used in manufacturing than any other machine tool. The essential components of a lathe are the bed, headstock assembly, tailstock assembly, and the leads crew and feed rod. The bed is the backbone of a lathe. It usually is made of well normalized or aged gray or nodular cast iron and provides s heavy, rigid frame on which all the other basic components are mounted. Two sets of parallel, longitudinal ways, inner and outer, are contained on the bed, usually on the upper side. Some makers use an inverted V-shape for all four ways, whereas others utilize one inverted V and one flat way in one or both sets, They are precision-machined to assure accuracy of alignment. On most modern lathes the way are surface-hardened to resist wear and abrasion, but precaution should be taken in operating a lathe to assure that the ways are not damaged. Any inaccuracy in them usually means that the accuracy of the entire lathe is destroyed. The headstock is mounted in a foxed position on the inner ways, usually at the left end of the bed. It provides a powered means of rotating the word at various speeds . Essentially, it consists of a hollow spindle, mounted in accurate bearings, and a set of transmission gears-similar to a truck transmission—through which the spindle can be rotated at a number of speeds. Most lathes provide from 8 to 18 speeds, usually in a geometric ratio, and on modern lathes all the speeds can be obtained merely by moving from two to four levers. An increasing trend is to provide a continuously variable speed range through electrical or mechanical drives. Because the accuracy of a lathe is greatly dependent on the spindle, it is of heavy construction and mounted in heavy bearings, usually preloaded tapered roller or ball types. The spindle has a hole extending through its length, through which long bar stock can be fed. The size of maximum size of bar stock that can be machined when the material must be fed through spindle. The tailsticd assembly consists, essentially, of three parts. A lower casting fits on the inner ways of the bed and can slide longitudinally thereon, with a means for clamping the entire assembly in any desired location, An upper casting fits on the lower one and can be moved transversely upon it, on some type of keyed ways, to permit aligning the assembly is the tailstock quill. This is a hollow steel cylinder, usually about 51 to 76mm(2to 3 inches) in diameter, that can be moved several inches longitudinally in and out of the upper casting by means of a hand wheel and screw. The size of a lathe is designated by two dimensions. The first is known as the swing. This is the maximum diameter of work that can be rotated on a lathe. It is approximately twice the distance between the line connecting the lathe centers and the nearest point on the ways, The second size dimension is the maximum distance between centers. The swing thus indicates the maximum work piece diameter that can be turned in the lathe, while the distance between centers indicates the maximum length of work piece that can be mounted between centers. Engine lathes are the type most frequently used in manufacturing. They are heavy-duty machine tools with all the components described previously and have power drive for all tool movements except on the compound rest. They commonly range in size from 305 to 610 mm(12 to 24 inches)swing and from 610 to 1219 mm(24 to 48 inches) center distances, but swings up to 1270 mm(50 inches) and center distances up

人脸识别文献综述

文献综述 1 引言 在计算机视觉和模式识别领域,人脸识别技术(Face Recognition Technology,简称FRT)是极具挑战性的课题之一。近年来,随着相关技术的飞速发展和实际需求的日益增长,它已逐渐引起越来越多研究人员的关注。人脸识别在许多领域有实际的和潜在的应用,在诸如证件检验、银行系统、军队安全、安全检查等方面都有相当广阔的应用前景。人脸识别技术用于司法领域,作为辅助手段,进行身份验证,罪犯识别等;用于商业领域,如银行信用卡的身份识别、安全识别系统等等。正是由于人脸识别有着广阔的应用前景,它才越来越成为当前模式识别和人工智能领域的一个研究热点。 虽然人类能够毫不费力的识别出人脸及其表情,但是人脸的机器自动识别仍然是一个高难度的课题。它牵涉到模式识别、图像处理及生理、心理等方面的诸多知识。与指纹、视网膜、虹膜、基因、声音等其他人体生物特征识别系统相比,人脸识别系统更加友好、直接,使用者也没有心理障碍。并且通过人脸的表情/姿态分析,还能获得其他识别系统难以获得的一些信息。 自动人脸识别可以表述为:对给定场景的静态或视频序列图像,利用人脸数据库验证、比对或指认校验场景中存在的人像,同时可以利用其他的间接信息,比如人种、年龄、性别、面部表情、语音等,以减小搜索范围提高识别效率。自上世纪90年代以来,人脸识别研究得到了长足发展,国内外许多知名的理工大学及TT公司都成立了专门的人脸识别研究组,相关的研究综述见文献[1-3]。 本文对近年来自动人脸识别研究进行了综述,分别从人脸识别涉及的理论,人脸检测与定位相关算法及人脸识别核心算法等方面进行了分类整理,并对具有典型意义的方法进行了较为详尽的分析对比。此外,本文还分析介绍了当前人脸识别的优势与困难。 2 人脸识别相关理论 图像是人们出生以来体验最丰富最重要的部分,图像可以以各种各样的形式出现,我们只有意识到不同种类图像的区别,才能更好的理解图像。要建立一套完整的人脸识别系统(Face Recognetion System,简称FRS),必然要综合运用以下几大学科领域的知识: 2.1 数字图像处理技术 数字图像处理又称为计算机图像处理,它是指将图像信号转换成数字信号并利用计算机

2019年我国人脸识别技术发展情况及发展趋势综合分析

2019年我国人脸识别技术发展情况 及发展趋势综合分析 2019年2月14日 一、全球生物识别细行业市场占比情况分析 生物识别指的是通过计算机与光学、声学、生物传感器和生物统计学原理等高科技手段相结合,利用人体固有的生理特性来进行个人身份鉴定技术。按不同的识别方式,生物识别可分为指纹识别、虹膜识别、语音识别、静脉识别和人脸识别。 伴随着生物识别产品逐渐从单一的PC处理转变为分布式计算。 用独立的前端独立设备来完成生物特征的采集、预处理、特征提取和比对,通过中心PC或服务器完成与业务相关的处理。随着生物特征 识别技术的不断发展和提高,生物特征识别技术的应用场景不断拓展,预计2015-2020年全球生物识别细分行业复合增长率分别为:人脸识别复合增长率为167%;语音识别为100%;虹膜识别为100%;指纹识别复合增长率为73%。

全球生物识别细行业市场占比情况 二、中国人脸识别技术发展情况分析 1、中国人脸识别行业发展历程 人脸识别技术在中国的发展起步于上世纪九十年代末,经历了技术引进-专业市场导入-技术完善-技术应用-各行业领域使用等五个阶段。其中,2014年是深度学习应用于人脸识别的关键一年,该年FaceBook发表一篇名为“DeepFace系统:达到肉眼级别的人脸识别系统”(翻译名),之后Face++创始人印奇团队以及香港中文大学汤晓鸥团队均在深度学习结合人脸识别领域取得优异效果,两者在LFW数据集上识别准确度均超过了99%,而肉眼在该数据集上的识别准确度仅为97.52%,可以说深度学习技术让计算机人脸识别能力超越人类的识别程度。

人脸识别与其他生物识别方式相比,优势在于自然性、不被察觉性等特点。自然性即该识别方式同人类进行个体识别时所利用的生物特征相同。指纹识别、虹膜识别等均不具有自然性。不被察觉的特点使该识别方法不易使人抵触,而指纹识别或虹膜识别需利用电子压力传感器或红外线采集指纹、虹膜图像,在采集过程中体验感不佳。目前人脸识别需要解决的难题是在不同场景、脸部遮挡等应用时如何保证识别率。此外,隐私性和安全性也是值得考虑的问题。 2、3D人脸识别与2D人脸识别数据对比 目前,国内的人脸识别技术已经相对发展成熟,该技术越来越多的被推广到安防领域,延伸出考勤机、门禁机等多种产品,产品系列达20多种类型,可以全面覆盖煤矿、楼宇、银行、军队、社会福利 保障、电子商务及安全防务等领域,人脸识别的全面应用时代已经到来。 中游人脸识别技术的进步,是推动下游场景应用拓展的关键所在。目前,人脸识别市场的解决方案主要包括2D识别、3D识别技术。市场上主流的识别方案是采用摄像头的2D方案,但由于人的脸部并非 平坦,因此2D识别在将3D人脸信息平面化投影的过程中存在特征信息损失。3D识别使用三维人脸立体建模方法,可最大程度保留有效 信息,因此3D人脸识别技术的算法比2D算法更合理并拥有更高精度。

论文《人工智能》---文献检索结课作业

人工智能 【摘要】:人工智能是一门极富挑战性的科学,但也是一门边沿学科。它属于自然科学和社会科学的交叉。涉及的学科主要有哲学、认知科学、数学、神经生理学、心理学、计算机科学、信息论、控制论、不定性论、仿生学等。人工智能(Artificial Intelligence),英文缩写为AI。它是研究、开发用于模拟、延伸和扩展人的智能的理论、方法、技术及应用系统的一门新的技术科学。人工智能是计算机科学的一个分支,它企图了解智能的实质,并生产出一种新的能以人类智能相似的方式做出反应的智能机器,该领域的研究包括机器人、语言识别、图像识别、自然语言处理和专家系统等1。 【关键词】:人工智能;应用领域;发展方向;人工检索。 1.人工智能描述 人工智能(Artificial Intelligence) ,英文缩写为AI。它是研究、开发用于模拟、延伸和扩展人的智能的理论、方法、技术及应用系统的一门新的技术科学2。人工智能是计 算机科学的一个分支,它企图了解智 能的实质,并生产出一种新的能以人 类智能相似的方式作出反应的智能 机器,该领域的研究包括机器人、语 言识别、图像识别、自然语言处理和 专家系统等。“人工智能”一词最初 是在1956 年Dartmouth学会上提出 的。从那以后,研究者们发展了众多 理论和原理,人工智能的概念也随之扩展。人工智能是一门极富挑战性的科学,从事这项工作的人必须懂得计算机知识,心理学和哲学。人工智能是包括十分广泛的科学,它由不同的领域组成,如机器学习,计算机视觉等等,总的说来,人工智能研究的一个主要目标是使机器能够胜任一些通常需要人类智能才能完成的复杂工作。但不同的时代、不同的人对这种“复杂工作”的理解是不同的。例如繁重的科学和工程计算本来是要人脑来承担的,现在计算机不但能完成这种计算, 而且能够比人脑做得更快、更准确,因之当代人已不再把这种计算看作是“需要人类智能才能完成的复 1.蔡自兴,徐光祐.人工智能及其应用.北京:清华大学出版社,2010 2元慧·议当人工智能的应用领域与发展状态〖J〗.2008

数控激光加工的技术水平以及应用外文文献翻译、中英文翻译

英文文献 Numerical control laser processing technology and application Pick to: as the main symbol of science and Technology development in the 20th century and the modern information society one of the pillars of optoelectronic Technology, Laser Technology, Laser Technology) and the development of Laser industry attaches great importance to by the advanced countries in the world. Numerical control and integrated the Laser and the optical system of the computer numerical control technology, advanced and high precision and automation of the workpiece positioning, the combination of formation, development and production of machining center has become a Laser machining, Laser Processing) is an important trend of development. Key words: CNC; Laser machining, Laser Processing); Level; application Before: laser is regarded as one of the most important scientific discoveries of the twentieth century, as soon as it was caused the great attention of materials scientists. In November 1971, gm pioneered the use of a 250 w CO2 laser to improve wear resistance materials using laser radiation test and research, and in 1974 successfully completed automobile redirector shell

人脸识别实验报告解读

人脸识别——特征脸方法 贾东亚12346046 一、实验目的 1、学会使用PCA主成分分析法。 2、初步了解人脸识别的特征法。 3、更熟练地掌握matlab的使用。 二、原理介绍 1、PCA(主成分分析法介绍) 引用一个网上的例子。假设有一份对遥控直升机操作员的调查,用表示飞行员i的 飞行技能,表示飞行员i喜欢飞行的程度。通常遥控直升飞机是很难操作的,只有那些 非常坚持而且真正喜欢驾驶的人才能熟练操作。所以这两个属性和相关性是非常强的。我们可以假设两者的关系是按正比关系变化的。如下图里的任意找的向量u1所示,数据散布在u1两侧,有少许噪声。 现在我们有两项数据,是二维的。那么如何将这两项变量转变为一个来描述飞行员呢?由图中的点的分布可知,如果我们找到一个方向的U,所有的数据点在U的方向上的投影之和最大,那么该U就能表示数据的大致走向。而在垂直于U的方向,各个数据点在该方向 的投影相对于在U上的投影如果足够小,那么我们可以忽略掉各数据在该方向的投影,这 样我们就把二维的数据转化成了在U方向上的一维数据。 为了将u选出来,我们先对数据进行预处理。先求出所有数据的平均值,然后用数据与平均值的偏差代替数据本身。然后对数据归一化以后,再代替数据本身。 而我们求最大的投影和,其实就是求各个数据点在U上的投影距离的方差最大。而X T u 就是投影的距离。故我们要求下式的最大值: 按照u是单位向量来最大化上式,就是求的特征向量。而此式是数据集的协方差矩阵。

在实际应用中,我们不止面临二维的数据。因此不能使用几何的形式呈现,但原理也是一样。就是找到一组相互正交的单位向量,然后根据贡献率考虑选择其中的部分作为考量的维数,这也就实现了数据的降维。 三、实验步骤 1、将库里的400张照片分成两组。一组作为训练,一组作为库。每个人的前五张照片作为 训练,后五张作为库。训练的照片按照顺序的数字重命名。库的照片名字不变。 2、库照片处理。 ①将每一张库的照片转化成N维的向量。(库里的照片是112*92,故将转化成的矩阵按列或行展开,就是个10304维的向量)我们稍后要对如此多维的向量用PCA进行降维。然后把这些向量存入一个矩阵里。而我是将这200个向量以列的形式存在了矩阵里。 即 ,,, ②将这200个向量的每个元素相加起来求出平均值。再用Z里的每一个向量减去这个平均值得到每个的偏差。 平均值,每个向量的偏差 即最后 ,,, ③接下来我们就要针对这些预处理后的数据进行降维。我们要求的N个相互正交的向量就是协方差矩阵的特征向量,而对应的特征值就是各个向量所占的比重。但是Z是个10304*200的矩阵,那么就是个10304*10304的矩阵。使用matlab直接求其特征值与特征向量不太实际。 所以我们考虑一个简单的运算方法: 协方差矩阵的秩受到训练图像的限制:如果有N个训练样本,则最多有N? 1 个对应非零特征值的特征向量,其他的特征向量对应的特征值都是0。如果训练样本的数目比图像的维数低,则可以通过如下方法简化主成份的计算。 设 Z是预处理图像的矩阵,每一列对应一个减去均值图像之后的图像。则,协方差矩阵为,并且对S的特征值分解为

相关文档