圖像特征提取方法(編輯修改稿)
2025-05-04 23:05 本頁(yè)面
【文章內(nèi)容簡(jiǎn)介】 plot(x1,y1)。 %~~~~亦可將圖像量化為13份（110 11 12 13）畫(huà)出bar投影圖像 num=zeros(1,13)。 for i=1:NUM number=length(find(L==i))。 for j=1:10 if number==j num(j)=num(j)+1。 end end if number10 amp。 number=20 num(11)=num(11)+1。 end if number20 amp。 number=100 num(12)=num(12)+1。 end if number100 num(13)=num(13)+1。 end end EH=[]。 for i=1:13 EH=[EH num(i)]。 end x=1:13。 y=EH。 figure(3),title(39。bar投影直方圖39。), bar(x,y)。、基于PCA的圖像數(shù)據(jù)特征提取clear all% Define databasex=imread(39。39。)。figure(1)plot(x(:,1),x(:,2),39。ko39。)xlabel(39。Valve Position (%)39。)ylabel(39。Fermenter Temperature (C)39。)% Calculate the covariance of dataset x (relationship between Valve and% Temperature)covariance=cov(x)。% Singular Value Deposition x=T*E*P% E is the eignvalue of x。 U is orthnormal。 V=U39。.%[U E V]=svd(covariance)。%U=U39。 V=V39。[U E] = eig(covariance)。E = diag([E(2,2),E(1,1)])。 % arrange the Eignvalue in a decending orderU = [U(:,2),U(:,1)]。 % arrange the Eignvector to make it patable with Eignvalue.% Mean center the raw dataValvebar=x(:,1)mean(x(:,1))。Tempbar=x(:,2)mean(x(:,2))。xbar=[Valvebar,Tempbar]。% Principal axis rotation of the covariance matrix.z=U39。*xbar39。z=z39。%covz is the covariance matrix of principal ponents which is equal to E%matrix. The variance of transformed variable (z1 amp。 z2) will have variance% and respectively.covz=U39。*covariance*U。% The first column of U factor is and + with nearly equal value which% means the first principal ponent is related to variability which both% measurements have difference. The second column is both positive which% means the 2nd PC is concentrate on the varaiability of the mon between% the two.% scaling of PCS.for i=1:2Vs(:,i)=sqrt(E(i,i))*U(:,i)。Ws(:,i)=U(:,i)/sqrt(E(i,i))。end% after rescale: Vs39。*Vs=E。 V39。*COV*Vs=E^2。 Ws39。*Ws=inv(E)。 Ws39。*COV*Ws=Iy=Ws39。*xbar39。%define T2T2=diag(y39。*y)。% find the control limit of T2p=2。 %PC numbern=10。 % sample numberT2Upper=p*(n1)/(np)*finv(,p,np)。% draw the T2 control chartsfigure(2)plot(T2,39。ko39。)hold onplot(T2Upper*ones(1,13),39。k39。)xlabel(39。Sample Number39。)ylabel(39。T2 Score39。)hold off% Draw the control ellipse% s1^2*s2^2/(s1^2*s2^2s12^2)*[(x1x1mean)^2/s1^2+(x2x2mean)^2/s2^22*s12(% x1x1mean)(x2x2mean)/s1^2/s2^2]=T2^2(upper control limit)% S=covariance。% a=S(1)*S(4)/(S(1)*S(4)S(2)^2)。% b=xbar(:,1).^2/S(1)+xbar(:,2).^2/S(4)2*S(2)*xbar(:,1).*xbar(:,2)./S(1)/S(4)。% error detection% two new points is added to be detected which are [,]。[。24]。% plot two values with the original 10.Valve2=[,]。Temp2=[,]。figure(1)hold onplot(Valve2,Temp2,39。k*39。)hold off% plot the control charts for valve and temperature separately.figure(3)hold onplot(1:10,Valve,39。ko39。)plot(11:12,Valve2,39。k*39。)plot(0:13, mean(Valve)*ones(1,14),39。k39。)plot(0:13, (mean(Valve)std(Valve)*2)*ones(1,14),39。k39。)plot(0:13, (mean(Valve)+std(Valve)*2)*ones(1,14),39。k39。)xlabel(39。Sample Number39。)ylabel(39。Valve Position (%)39。)hold offfigure(4)hold onplot(1:10,Temp,39。ko39。)plot(11:12,Temp2,39。k*39。)plot(0:13, mean(Temp)*ones(1,14),39。k39。)plot(0:13, (mean(Temp)+std(Temp)*2)*ones(1,14),39。k39。)plot(0:13, (mean(Temp)std(Temp)*2)*ones(1,14),39。k39。)xlabel(39。Sample Number39。)ylabel(39。Temperature (C)39。)hold off% transform new observations to PC axisy2=Ws39。*([Valve2mean(Valve)。Temp2mean(Temp)])。% plot the control charts for principal ponentsfigure(5)hold onplot(1:10,y(1,:),39。ko39。)plot(11:12,y2(1,:),39。k*39。)plot(0:13, mean(y(1,:))*ones(1,14),39。k39。)plot(0:13, 2*ones(1,14),39。k39。)plot(0:13, 2*ones(1,14),39。k39。)xlabel (39。Sample Number39。)ylabel (39。Principal Component 139。)hold offfigure(6)hold onplot(1:10,y(2,:),39。ko39。)plot(11:12
公司管理相關(guān)推薦
鄂ICP備17016276號(hào)-1