【正文】
現(xiàn)有的識別方法中,通過從人臉圖像中提取出特征信息,來對數(shù)據(jù)庫進(jìn)行檢索的方法速度快,而利用拓?fù)鋵傩詧D匹配來確定匹配度的方法則相對較快。人臉識別早在六七十年代就引起了研究者的強(qiáng)烈興趣。endx = checklimit(endx,limx)。 % 調(diào)整模板大小model_rot = imrotate(model_rot,angle,39。 end。for j=1:n, for i=1:m, if (A(i,j) ~= 0) left = j。 % 將 uint8 型轉(zhuǎn)換成 double 型G1=im2double(G)。在對大量參考文獻(xiàn)資料的閱讀的基礎(chǔ)上,本設(shè)計(jì)對基于 Matlab 的人臉識別這一技術(shù)做了詳細(xì)的綜述。,OutputName)。*.bmp39。},39。% Executes on button press in pushbutton1.function pushbutton1_Callback(hObject, eventdata, handles)% hObject handle to pushbutton1 (see GCBO)% eventdata reserved to be defined in a future version of MATLAB% handles structure with handles and user data (see GUIDATA)global TrainDatabasePath 。gui_OutputF39。人臉識別,即通過對所采集到的人臉圖像進(jìn)行一系列處理,提取待識別人臉圖像的特征信息,通過與已存人臉數(shù)據(jù)庫信息進(jìn)行匹配識別,確定待識別人臉圖像的基本信息。BB2=cell2mat(BB1)。 if pr=100 BW(x1:x2, y1:y2)=0。以上四種方法的優(yōu)缺點(diǎn)比較見表 32:表 32 優(yōu)缺點(diǎn)對比表檢測方法 優(yōu) 點(diǎn) 缺 點(diǎn)神經(jīng)網(wǎng)絡(luò) 效率較高,錯(cuò)誤報(bào)警數(shù)較少,網(wǎng)絡(luò) 多樣本訓(xùn)練所耗的費(fèi)時(shí)間多,網(wǎng)絡(luò)監(jiān)測檢測方法 優(yōu) 點(diǎn) 缺 點(diǎn)模板匹配具有較強(qiáng)的直觀性和較好的適應(yīng)性對面部表情的變換敏感;對于模板的選擇、參數(shù)的確定很困難膚色模型 檢測速度相對較快陽光、背景光線等會(huì)使人臉區(qū)域被分割,導(dǎo)致被漏檢先驗(yàn)知識的方法對于復(fù)雜圖像中的人臉檢測有較大優(yōu)勢依賴于先驗(yàn)知識;工作量較大,運(yùn)算時(shí)間較長14法 監(jiān)測速度較快 錯(cuò)誤報(bào)警數(shù)較多本征臉法能抽象人臉全部信息,運(yùn)算時(shí)間相對較短通過模板測效率較低,多模板雖然增加了效率,但是檢測時(shí)間較長積分圖像分析法檢測速度較快,滿足實(shí)時(shí)檢測的要求,檢測效率相對較高錯(cuò)誤報(bào)警數(shù)與檢測率成反比支撐向量法具有更好的泛化能力“非人臉”的復(fù)雜造成支持向量數(shù)目較多,導(dǎo)致運(yùn)算復(fù)雜度變大運(yùn)用 matlab 軟件仿真進(jìn)行人臉檢測定位實(shí)例:人臉檢測定位程序:%%%%% Reading of a RGB image原始圖像 i=imread(39。 圖像處理工具箱中提供了 edge()函數(shù)來實(shí)現(xiàn)圖像邊緣檢測,還有各種方法算子供我們選擇,在本案例中采用了 canny 算子來進(jìn)行圖像邊緣檢測,程序代碼如下:i=imread(39。)。例如,消除照片中的劃痕,改善光照不均勻圖像,突出目標(biāo)的邊緣等。 人臉圖像的讀取與顯示人臉圖像的讀取和顯示可通過 imread( )和 imshow( )指令來實(shí)現(xiàn)。5第 2 章 圖像處理的 Matlab 實(shí)現(xiàn) 識別系統(tǒng)構(gòu)成人臉識別技術(shù)系統(tǒng)主要可分為四個(gè)組成部分,分別為:人臉圖像采集及檢測、人臉圖像預(yù)處理、人臉圖像特征提取以及匹配與識別。它是人們一直所追求的讓機(jī)器智能化技術(shù),就是讓機(jī)器具備和人類一樣的思考能力,識別能力以及處理事務(wù)的能力。采用快速人臉檢測識別技術(shù)可以從視頻監(jiān)控圖象中實(shí)時(shí)捕獲到人臉信息,并與人臉數(shù)據(jù)庫中的已存信息進(jìn)行實(shí)時(shí)比對,從而達(dá)到快速身份識別的效果。目前,人臉識別技術(shù)應(yīng)用最廣泛的地方就是各大公司、商場、政府保密機(jī)構(gòu)的門禁考勤系統(tǒng)。鐵路部門發(fā)布計(jì)劃時(shí)表示,將在京滬高鐵段的天津西站、濟(jì)南西站、上海虹橋站這三個(gè)站點(diǎn),建立人臉識別系統(tǒng)工程,以此來協(xié)助公安部門甄別、抓捕在逃罪犯。人臉檢測就是挑出這其中有用的特征信息,并利用這些特征來實(shí)現(xiàn)人臉識別。在類型轉(zhuǎn)換的處理過程中,我們還會(huì)經(jīng)常遇到數(shù)據(jù)類型不匹配的問題,針對這一問題, 工具箱中為我們提供了各種數(shù)據(jù)類型之間相互轉(zhuǎn)換的函數(shù),例如 double()函數(shù)的功能就是將數(shù)據(jù)轉(zhuǎn)換為雙精度數(shù)據(jù)類型。實(shí)現(xiàn)過程代碼如下:8i=imread(39。j1=wiener2(j)。,[,],)。figure,imshow(BW) %%%%%%%%%%%%%%%%%%%%灰度圖像及均衡化灰度圖像 [n1 n2]=size(BW)。 pr1=0。 (BB2(1,k)/BB2(1,k+1)) mx=p。,39。, [])。 )。%axes()。im=imread(str)。Equivalent Image39。在做畢業(yè)設(shè)計(jì)的這段時(shí)間里,我的老師、同學(xué)們對我給予了非常多的幫助,在這里,謹(jǐn)向他們致以最真誠的感謝!尤為感謝的,是我的導(dǎo)師周經(jīng)國老師。 % 列像素for i=1:row28for j=1:column rr(i,j)=R1(i,j)/RGB(i,j)。 end。for i=m:1:1, for j=1:n, if (A(i,j) ~= 0) down = i。 % 選擇模板人臉區(qū)域[modx,mody] =center(bwmodel_rot)。ccorr = corr2(mfit,mult) % 計(jì)算相關(guān)度[l,r,u,d] = bianjie(bwmodel_rot)。80 年代初 T. Minami 研究出了優(yōu)于 Sakai 的人臉圖像自動(dòng)識別系統(tǒng)。什么時(shí)候離光明最近?那就是你覺得黑暗太黑的時(shí)候。國外有許多學(xué)校在研究人臉識別技術(shù),研究涉及的領(lǐng)域很廣。t get rid of the intervention of people. Into the niy s, as a result of the parties face the pressing needs of the face recognition system, the research of face recognition is very popular. Face recognition method is a major breakthrough, entered the stage of the real machine automatic identification such as Karhunen Loeve transform or a new neural work technology. Face recognition research obtained the unprecedented attention, the number of papers published on face recognition and so on increased dramatically, from 1990 to 1990 alone, between SCI and EI can be retrieved as many as thousands of articles, related literature about face recognition during the period of this review is also visible. Abroad, there are many schools in facial recognition technology research, research field is very wide. These studies are military, police and big pany attaches great importance to and support the domestic some wellknown colleges and universities are engaged in the research of face recognition.Face recognition is a frontier topic in the field of pattern recognition, but the current face recognition was still in the stage of research topic, it is not active topic in the field of practical application. While humans can tell a person had no difficulty (in) the human face, but the use of puters for fully automatic face recognition has many difficulties, displays in: the face is a rigid body, the face changes。starty = cymody。end。if (right ~= 1) break。(2)找區(qū)域邊界function [left, right, up, down] = bianjie(A)[m n] = size(A)。最后,我還要感謝身邊的朋友和同學(xué),在大學(xué)生活的四年里,我在你們的陪伴中成長,謝謝你們在做畢業(yè)設(shè)計(jì)這段日子里給予我的幫助。目前到處可見攝像頭,監(jiān)控錄像,這些的普及,使人臉識別具有重大商業(yè)價(jià)值。global TrainDatabasePath 。%[m V_PCA V_Fisher ProjectedImages_Fisher] = FisherfaceCore(T)。)。19end% End initialization code DO NOT EDIT% Executes just before faceCore is made visible.function faceCore_OpeningF(hObject, eventdata, handles, varargin)% This function has no output args, see OutputF.% hObject handle to figure% eventdata reserved to be defined in a future version of MATLAB% handles structure with handles and user data (see GUIDATA)% varargin mand line arguments to faceCore (see VARARGIN)% Choose default mand line output for faceCore = hObject。, mfilename, ... 39。,[BB2(1,j2),BB2(1,j1),BB2(1,j),BB2(1,j)],39。end