使用MATLAB自带的PCA函数进行人脸识别_.zip

clear clc % 样本数量15*11 people_count=15; face_count_per_people=11; % 训练比率,70%时识别正确率为96.67%,设置为75%识别正确率可达100% training_ratio=.75; % 能量 energy=90; training_count=floor(face_count_per_people*training_ratio); training_samples=[]; path_mask='face_rec\\%03d\\%02d.jpg'; % 训练 for i=1:people_count for j=1:training_count img=im2double(imread(sprintf(path_mask,i,j))); img=imresize(img,[10 10]); % 归一化至50*50 if ndims(img)==3 img=rgb2gray(img); end training_samples=[training_samples;img(:)']; end end mu=mean(training_samples); [coeff,scores,~,~,explained]=pca(training_samples); idx=find(cumsum(explained)>energy,1); coeff=coeff(:,1:idx); scores=scores(:,1:idx); % 测试 acc_count=0; for i=1:people_count for j=training_count+1:face_count_per_people img=im2double(imread(sprintf(path_mask,i,j))); img=imresize(img,[10 10]); if ndims(img)==3 img=rgb2gray(img); end score=(img(:)'-mu)/coeff'; [~,idx]=min(sum((scores-repmat(score,size(scores,1),1)).^2,2)); if ceil(idx/training_count)==i acc_count=acc_count+1; end end end test_count=(people_count*(face_count_per_people-training_count)); acc_ratio=acc_count/test_count; fprintf('测试样本数量:%d,正确识别率:%2.2f%%',test_count,acc_ratio*100)