《智能檢測(cè)技術(shù)》教學(xué)參考材料(王偉版)附錄主成分分析、一元線性回歸、多元線性回歸、主成分回歸、偏最小二乘回歸、支持向量機(jī)、偏最小二乘判別、車牌照識(shí)別等算法的MATLAB實(shí)現(xiàn)代碼如下所示：主成分分析：.功能：將高維數(shù)據(jù)映射到低維的空間中，并期望在所投影的維度上數(shù)據(jù)的信息量最大(方差最大)，以此使用較少的數(shù)據(jù)維度，同時(shí)保存住較多的原數(shù)據(jù)點(diǎn)。.注解：hald為matlab自帶數(shù)據(jù)集；pea函數(shù)為主成分分析核心函數(shù)，其中輸入變量ingredients為nxp矩陣，n為觀測(cè)樣本數(shù)，p為變量數(shù)，輸出變量coeff為主成分系數(shù)，score為主成分分?jǐn)?shù)，latent為主成分方差，tsquared為輸入變量的每個(gè)觀測(cè)值的HotellingT方統(tǒng)計(jì)量，explained為每個(gè)主成分解釋的總方差百分比，mu為ingredients中每個(gè)變量的估計(jì)均值；latentO中存放每個(gè)主成分對(duì)應(yīng)貢獻(xiàn)值；繪制主成分分類圖。.代碼：loadhald[coeff,score,latent,tsquared,explained,mu]=pca(ingredients);latentO1OO*latent/sum(latent);figure;scatter3(score(1:4,1),score(1:4,2),score(l:4,3),100,V,Tilled1);holdonscatter3(score(5:8,1),score(5:8,2),score(5:8,3),100,Tilled');holdonscatter3(score(9:13,1),score(9:13,2),score(9:13,3),100,'g；'s；'filled');holdoff一元線性回歸：.功能：由兩個(gè)相關(guān)變量中的一個(gè)變量X估計(jì)另一個(gè)變量y,建立最優(yōu)的數(shù)學(xué)模型y=/(%)。.注解：實(shí)際問(wèn)題引入:社會(huì)商品銷售總額與職工工資總額，建立其回歸模型，設(shè)x為職工工資總額，y為商品銷售總額，建立x—y模型；Linear.fit為擬合函數(shù)，m2為建立模型，wlb為方差分析表，y_new為預(yù)測(cè)值，y_newr為預(yù)測(cè)區(qū)間。[r,s]=size(CuDingWeiErZhi);YuJingDingWei=double(CuDingWeiErZhi);X2=zeros(l,s);fori=l:rforj=l:sif(YuJingDingWei(i,j)==1)X2(l,j)=X2(lj)+1;endendend[temp,MaxX]=max(X2);subplot(2,2,2),plot(0:s-l,X2),titleC粗定位車牌圖像列方向像素點(diǎn)值累計(jì)和)xlabelC列值)ylabel(像素)；[g,h]=size(YuJingDingWei);ZuoKuanDu=0;YouKuanDu=0;KuanDuYuZhi=5;whilesum(YuJingDingWei(:,ZuoKuanDu+1))—0ZuoKuanDu=ZuoKuanDu4-1;endifZuoKuanDu<KuanDuYuZhiYuJingDingWei(:,[l:ZuoKuanDu])=0;YuJingDingWei=QieGe(YuJingDingWei);endsubplot(2,2,3),imshow(YuJingDingWei),titleC去除左側(cè)邊框的二值車牌圖像')[e,f]=size(YuJingDingWei);d=f;whilesum(YuJingDingWei(:,d-l))^-0YouKuanDu=￥ouKuanDu+1;d=d-l;endifYouKuanDu<KuanDuYuZhiYuJingDingWei(:,[(f-YouKuanDu):f])=0;YuJingDingWei=QieGe(YuJingDingWei);endsubplot(2,2,4),imshow(YuJingDingWei),title，精確定位的車牌二值圖像)ChePaiErZhi=YuJingDingWei;%logical()ChePaiLvBo=bwareaopen(ChePaiErZhi,20);subplot(l,2,1),imshow(ChePaiLvBo),title(形態(tài)學(xué)濾波后的車牌二值圖像)ChePaiYuFenGe=double(ChePaiLvBo);|p,q]=size(ChePaiYuFenGe);X3=zeros(l,q);forj=l:qfori=l:pif(ChePaiYuFenGe(i,j)==1)X3(lj)=X3(l,j)+l;endendendsubplot(1,2,2),plot(0:q-1,X3),titled列方向像素點(diǎn)灰度值累計(jì)和)xlab&C列值)ylabelC累計(jì)像素量)PxO=q;Pxl=q;fori=l:6while((X3(1,Px0)<3)&&(Px0>0))PxO=PxO-l;endPxl=PxO;while(((X3(l,Px1)>=3))&&(Px1>0)||((Px0-Px1)<15))Px1=Px1-1;endChePaiFenGe=ChePaiLvBo(:,Pxl:PxO,:);figure(6);subplot(l,7,8-i);imshow(ChePaiFenGe);ii=int2str(8-i);imwrite(ChePaiFenGe,strcat(ii,,.jpg,));PxO=Px1;endPX3=Pxl;while((X3(1,PX3)<3)&&(PX3>0))PX3=PX3-1;endZiFulDingWei=ChePaiYuFenGe(:,1:PX3,:);subplot(1,7,1);imshow(ZiFuIDingWei);imwrite(ZiFu1DingWei,4.jpg');ZiFul=imresize(-imread(,l.jpg1),[11055],'bilinear');ZiFu2=imresize(-imread(,2.jpg,),[11055],'bilinear');ZiFu3=imresize(-imread(,3.jpg,),[11055],'bilinear');ZiFu4=imresize(-imread(,4.jpg,),[11055],'bilinear');ZiFu5=imresize(-imread(,5.jpg,),[11055],'bilinear');ZiFu6=imresize(-imread(,6.jpg'),[11055],'bilinear');ZiFu7=imresize(^imread(,7.jpg,),[11055],'bilinear');HanZi=DuQuHanZi(imread(,MuBanKu\sichuan.bmp,),imread(,MuBanKu\guizhou.bmp,),imread(,MuBanKu\beijing.bmp'),imread(,MuBanKu\chongqing.bmp,),...imread('MuBanKu\guangdong.bmp'),imreadCMuBanKu\shandong.bmp'),imreadCMuBanKu\zhejiang.bmp'))；ShuZiZiMu=DuQuSZZM(imread('MuBanKu\0.bmp'),imread('MuBanKu\l.bmp'),imread('MuBanKu\2.bmp,),imread(,MuBanKu\3.bmp,),imread(,MuBanKu\4.bmp,),...imread('MuBanKu\5.bmp'),imread('MuBanKu\6.bmp'),imread('MuBanKu\7.bmp'),imread('MuBanKu\8.bmp,),imread(,MuBanKu\9.bmp,),...imread('MuBanKu\10.bmp'),imread('MuBanKu\l1.bmp'),imread('MuBanKu\l2.bmp'),imread('MuBanKu\l3.bmp'),imreadCMuBanKu\14.bmp'),…imread(,MuBanKu\15.bmp,),imread(,MuBanKu\16.bmp,),imread(,MuBanKu\17.bmp,),imread(,MuBanKu\l8.bmp1),imread(*MuBanKu\l9.bmp'),...imread('MuBanKu\20,bmp'),imread('MuBanKu\21.bmp'),imreadCMuBanKu\22.bmp'),imread('MuBanKu\23.bmp'),imread('MuBanKu\24.bmp'),??.imread('MuBanKu\25,bmp'),imread('MuBanKu\26.bmp'),imread('MuBanKu\27,bmp'),imread('MuBanKu\28.bmp'),imreadCMuBanKu\29.bmp'),…imread(,MuBanKu\30.bmp,),imread('MuBanKu\31.bmp'),iniread(,MuBanKu\32.bmp,),imread(,MuBanKu\33.bmp'));ZiMu=DuQuZiMu(imread('MuBanKu\10.bmp'),imread('MuBanKu\ll.bmp'),imread('MuBanKu\12?bmp^imreadCMuBanKuM3.bmp1),imreadCMuBanKu\14.bmp'),...imread(,MuBanKu\15.bmp,),imread('MuBanKu\l6.bmp*),imreadC^uBanKuM7.bmp1),imreadC/MuBanKuM8.bmp'),imreadC^uBanKuM9.bmp'),...imread(,MuBanKu\20.bmp,),imread(,MuBanKu\21.bmp'),iniread(,MuBanKu\22.bmp,),imread(,MuBanKu\23.bmp'),imreadCMuBanKu\24.bmp'),?..imread(,MuBanKu\25.bmp,),imread(,MuBanKu\26.bmp'),imread(,MuBanKu\27.bmp'),imread(,MuBanKu\28.bmp'),imreadCMuBanKu\29.bmp'),…imread('MuBanKu\30,bmp'),imread('MuBanKu\31.bmp'),imread('MuBanKu\32.bmp'),imread('MuBanKu\33.bmp*));ShuZi=DuQuShuZi(imread('MuBanKu\0.bmp'),imread('MuBanKu\l.bmp'),imread('MuBanKu\2.bmp'),imread(,MuBanKu\3.bmp,),imreadCMuBanKu\4.bmp'),...imread(,MuBanKu\5.bmp,),imread(,MuBanKu\6.bmp,),imread(,MuBanKu\7.bmp,),imread(,MuBanKu\8.bmp^imreadCMuBanKu\9.bmp1));t=l；ZiFuUieGuo=ShiBieHanZi(HanZi,ZiFu1);ShiBieJieGuo(l,t)=ZiFulJieGuo;t=t+1;ZiFu2JieGuo=ShiBieZiMu(ZiMu,ZiFu2);ShiBieJieGuo(l,t)=ZiFu2JieGuo;t=t+l;ZiFu3JieGuo=ShiBieSZZM(ShuZiZiMu,ZiFu3);ShiBieJieGuo(l,t)=ZiFu3JieGuo;t=t+l;ZiFu4JieGuo=ShiBieSZZM(ShuZiZiMu,ZiFu4);ShiBieJieGuo(l,t)=ZiFu4JieGuo;t=t+1；ZiFu5JieGuo=ShiBieShuZi(ShuZi,ZiFu5);ShiBieJieGuo(1,t)=ZiFu5JieGuo;t=t+1;ZiFu6JieGuo=ShiBieShuZi(ShuZi,ZiFu6);ShiBieJieGuo(1,t)=ZiFu6JieGuo;t=t+1;ZiFu7JieGuo=ShiBieShuZi(ShuZi,ZiFu7);ShiBieJieGuo(1,t)=ZiFu7JieGuo;t=t+1;ShiBieJieGuomsgbox(ShiBieJieGuo；結(jié)果)；fid=fopen('Data.xls7a+,);fprintf(fid/%s\r\n',ShiBieJieGuo,datestr(now));fclose(fid);.代碼：x=[23.80,27.60,31.60,32.40,33.70,34.90,43.20,52.80,63.80,73.40];y=[41.4,51.8,61.70,67.90,68.70,77.50,95,90,137.40,155.0,175.0];figureplot(x,y,T*')xlabel('x(職工工資總額)「fontsize',12)ylabelCy(商品零售總額)「fontsize；12)set(gca,linewidth1,2)Lxx=sum((x-mean(x)).A2);Lxy=sum((x-mean(x)).*(y-mean(y)));bl=Lxy/Lxx;b0=mean(y)-b1*mean(x);yl=bl*x+bO;holdonplot(x,yl,linewidth1,2);m2=LinearModel.fit(x,y);m2.plotwlb=m2.anova;Xnew=[23.8;63.8];[y_new,y_newr]=m2.predict(Xnew)多元線性回歸：.功能：因變量的變化往往受幾個(gè)重要因素的影響，并且自變量(Xi，X2,…,Xn)與因變量(y)之間為線性關(guān)系時(shí)一，建立多元線性回歸模型y=b0+瓦％]4-b2x2+…+bnxn,其中外為常數(shù)項(xiàng)，瓦,歷,…,一為回歸系數(shù)。.注解：假設(shè)計(jì)算具有常數(shù)項(xiàng)(截距)的模型的系數(shù)估計(jì)值，請(qǐng)?jiān)诰仃嘪中包含一個(gè)由1構(gòu)成的列。regress函數(shù)為多元線性回歸核心函數(shù)，其中輸出參數(shù)：b為多元線性回歸的系數(shù)估計(jì)值，bint為系數(shù)估計(jì)值的置信邊界下限和置信邊界上限，r為由殘差組成的向量，rint為可用于診斷離群值的區(qū)間，stats為模型統(tǒng)計(jì)量.包含R2統(tǒng)計(jì)量、F統(tǒng)計(jì)量及其p值，以及誤差方差的估計(jì)值，輸入?yún)?shù)：y為相應(yīng)矩陣，nXl數(shù)值向量，x為預(yù)測(cè)變量數(shù)據(jù)，nXp數(shù)值矩陣。.代碼:clc;clear;y=[7613.517850.918381.869142.8110813.68631.438124.949429.7910230.8110163.619737.568561.067781.827110.97];xl=[76667704814885718679770464715870528938153335292727582591];x2=[16.2216.8517.9317.2817.23171918.2216.313.3711.6210.369.839.25];X=[ones(size(y));xl.A2;x2.A2;xl;x2;xl.*x2];[b,bint,r,rint,stats]=regress(y',X');scatter3(xl,x2,y,Tilled')holdonxlfit=min(xl):100:max(xl);x2fit=min(x2):1:max(x2);[X1FIT,X2HT]=meshgrid(xlfit,x2fit);YFIT=b(l)+b(2)*X1FIT.A2+b(3)*X2FIT.A2+b(4)*X1FIT+b(5)*X2FIT+b(6)*XlFIT.*X2FIT;mesh(XlFIT,X2FIT,YFIT)view(10,10)xlabel(!xr)ylabel('x2')zlabel('y')主成分回歸：.功能：對(duì)回歸模型中的多重共線性進(jìn)行消除后，將主成分變量作為自變量進(jìn)行回歸分析，然后根據(jù)得分系數(shù)矩陣將原變量代回得到的新的模型。.注解：NIR為60個(gè)汽油樣本的401個(gè)波長(zhǎng)的光頻譜強(qiáng)度；octane為對(duì)應(yīng)的辛烷值；PCRmsep為交叉驗(yàn)證選擇主成分?jǐn)?shù)；rsquaredPCR為R方值。.代碼：loadspectra[dummy,h]=sort(octane);oldorder二get(gcf,BefaultAxesColorOrder1);set(gcf,'DefaultAxesColorOrdef,jet(60));plot3(repmat(1:401,60,1)\repmat(octane(h),1,401),,NIR(h,:)1);set(gcf,'DefaultAxesColorOrder；oldorder);xlabel(*WavelengthIndex');ylabel('Octane');axis('tight');gridonX=NIR;y=octane;[n,p]=size(X);[PCALoadings,PCAScores,PCAVar]=pca(X,'Economy',false);PCRmsep=sum(crossval(@pcrsse,X,y,KFold1,10),1)/n;plot(0:10,PCRmsep,fr-A，);xlabel(*Numberofcomponents1);ylabel('EstimatedMeanSquaredPredictionError1);legendCPCRVlocationVNE1);betaPCR=regress(y-mean(y),PCAScores(:,l:4));betaPCR=PCALoadings(:,l:4)*betaPCR;betaPCR=[mean(y)-mean(X)*betaPCR;betaPCR];yfitPCR=[ones(nj)X]*betaPCR;plot(y,yfitPCR,YA，);xlabel('ObservedResponse1);ylabel('FittedResponse');legend({*PCRwith4Components1),location1,'NW1);TSS=sum((y-mean(y)).A2);RSS_PCR=sum((y-yfitPCR).A2);rsquaredPCR=1-RSS_PCR/TSS;plot(1:401,PCALoadings(:,1xlabel(,Variable,);ylabel('PCALoading1);legend({'lstComponent1'2ndComponent'3rdComponent'...'4thComponent'}[location；'NW');偏最小二乘回歸：.功能：查找兩個(gè)矩陣(X和Y)的基本關(guān)系，試圖找到X空間的多維方向來(lái)解釋丫空間方差最大的多維方向。偏最小二乘回歸特別適合當(dāng)預(yù)測(cè)矩陣比觀測(cè)的有更多變量，以及X的值中有多重共線性的情況。.注解：加載matlab數(shù)據(jù)spectra,plsrgress函數(shù)中參數(shù)CV=10為10折交叉驗(yàn)證，rsquaredPLS為決定系數(shù)，plsrgress函數(shù)輸入?yún)?shù)中，X為預(yù)測(cè)變量，丫為相應(yīng)變量，ncomp為主成分?jǐn)?shù)，輸出參數(shù)中XL為預(yù)測(cè)載荷，YL為相應(yīng)載荷，XS為預(yù)測(cè)分?jǐn)?shù)，YS為反響分?jǐn)?shù)，beta為PLS回歸的系數(shù)估計(jì)，pctVar為方差百分比，PLSmsep為均方誤差。.代碼：loadspectra[dummy,h]=sort(octane);oldorder=get(gcf,^efaultAxesColorOrder*);set(gcf,'DefaultAxesColorOrdef,jet(60));plot3(repmat(l:401,60,l)\repmat(octane(h),1,401)\NIR(h,:)t);set(gcf,'DefaultAxesColorOrder；oldorder);xlabel(!WavelengthIndex');ylabelCOctane1);axis('tightf);gridonX=NIR;y=octane;[n,p]=size(X);[XI,YI,Xs,Ys,beta,pctVar,PLSmsep]=plsregress(X,y,10,'CV；10);plot(0:10,PLSmsep(2,:),'b-o');xlabel(fNumberofcomponents');ylabel('EstimatedMeanSquaredPredictionError*);legendCPLSR；location1,NE);[Xloadings,Yloadings,Xscores,Yscores,betaPLS]=plsregress(X,y,3);yfitPLS=|ones(n,l)X]*betaPLS;plot(y,yfitPLS,'bo*);xlabel('ObservedResponse1);ylabel('FittedResponse');legend({PLSRwith3Components'},'location*,,NW1);TSS=sum((y-mean(y)).A2);RSS_PLS=sum((y-yfitPLS).A2);rsquaredPLS=1-RSS_PLS/TSS;[XI,Yl,Xs,Ys,beta,pctVar,mse,stats]=plsregress(X,y,3);plot(l:401,stats.W/-1);xlabel(!Variable1);ylabel(PLSWeight1);legend({'lstComponent*'2ndComponent'3rdComponent'),'location*,'NW1);支持向量機(jī)：.功能：通過(guò)別離超平面把原始樣本集進(jìn)行分類。.注解：加載Fisher的虹膜數(shù)據(jù)集，將Y中的80%作為訓(xùn)練集，20%為測(cè)試集，fitcsvm為模型訓(xùn)練，predict為預(yù)測(cè)函數(shù)，errRate為分類錯(cuò)誤率，conMat為混淆矩陣。.代碼：loadfisheririsX=[meas(:,1),meas(:,2)];Y=nominal(ismember(species,^etosa'));P=cvpartition(Y,'Holdout1,0.20);svmStruct=fitcsvm(X(P.training,:),Y(P.training));C=predict(svmStruct,X(P.test,:));errRate=sum(Y(P.test)-=C)/P.TestSize;conMat=confusionmat(Y(P.test),C);偏最小二乘判別：.功能：用于判別分析的多變量統(tǒng)計(jì)分析方法，可以減少變量間多重共線性產(chǎn)生的影響.注解：加載紅酒數(shù)據(jù)，并劃分訓(xùn)練集與驗(yàn)證集，交叉驗(yàn)證選擇最優(yōu)主成分，plsdafit為模型訓(xùn)練函數(shù)，model為建立的模型，accuracy_train為訓(xùn)練集準(zhǔn)確率，accuracy_test為測(cè)試集準(zhǔn)確率。.代碼：clear;clc;[data,label]=wine_dataset;labels=mod(find(laber==1),3)；data二data';num_test=78;[ndata,D]=size(data);R=randperm(ndata);test_X=data(R(l:num_test),:);test_Y=labels(R(1:num_test),:);test_Y=double(test_Y);R(l:num_test)=[];train_X=data(R,:);num_train=size(train_X,l);train_Y=labels(R,:);train_Y=double(train_Y);res=plsdacompsel(train_X,train_Y,'meancentering'/venetianblinds1,10,'max');accuracy_cv=max(res.ner);comp=find(res.ner=max(res.ner));model=plsdafit(train_X,train_Y,comp(1,1),'meancentering7max\l);accuracy_train=model.class_param.accuracy;pred=plsdapred(test_X,model);test_result」abel二pred.class_pred;test_param=calc_class_param(test_result_label,test_Y);accuracy_test=test_param.accuracy;車牌號(hào)碼識(shí)別：.功能：實(shí)現(xiàn)對(duì)含有車牌的圖片進(jìn)行自動(dòng)車牌號(hào)碼提取。.注解：首先進(jìn)行車牌區(qū)域粗定位，對(duì)車牌區(qū)域進(jìn)行字符分割，并使用字符數(shù)字模板庫(kù)進(jìn)行車牌號(hào)碼識(shí)別，實(shí)現(xiàn)車牌號(hào)碼檢測(cè)。.代碼：closeall;clc;[fn,pn,fi]=uigetfHe('ChePaiKu\*.jpg?選擇圖片')；YuanShi=imread([pnfn]);figure(1);subplot(3,2,1),imshow(YuanShi),title('原始圖像')；YuanShiHuiDu=rgb2gray(YuanShi);subplot(3,2,2),imshow(YuanShiHuiDu),title，灰度圖像，)；BianYuan=edge(YuanShiHuiDu,'canny',0.5);subplot?,2,3),imshow(BianYuan),title(!Canny算子邊緣檢測(cè)后圖像')；sel=[l;l;l];FuShi=imerode(BianYuan,sei);subplot(3,2,4),imshow(FuShi),t