人脸识别仿真,提取ORL人脸数据库的协方差矩阵S的特征值和特征向量,通过PCA降维后人脸识别-源码

clc; clear; close all; warning off; addpath(genpath(pwd)); image_dims = [112, 92]; num_images = 40; test_images = cell(40, 3); train_images = zeros(prod(image_dims), num_images); input_dir = 'att_faces'; for i=1:40 sub_dir = strcat('s', num2str(i)); images = cell(10); for j=1:10 filename = fullfile(input_dir, sub_dir, strcat(num2str(j), '.pgm')); images{j} = imread(filename); end images = images(randperm(10)); img = zeros(image_dims); for j=1:7 img = img + double(images{j}); end img = img / 7; train_images(:, i) = img(:); for j=8:10 test_images{i,j-7}=images{j}; end end % steps 1: find the mean image and the mean-shifted input images mean_face = mean(train_images, 2); shifted_images = train_images - repmat(mean_face, 1, num_images); % steps 2: calculate the ordered eigenvectors and eigenvalues [full_evectors, score, evalues] = pca(train_images'); result = zeros(39, 2); % step 4: only retain the top 'num_eigenfaces' eigenvectors (i.e. the principal components) for num_eigenfaces = 1:39 evectors = full_evectors(:, 1:num_eigenfaces); features = evectors' * shifted_images; cnt = 0; for i=1:40 for j=1:3 input_image = double(test_images{i,j}); % calculate the similarity of the input to each training image feature_vec = evectors' * (input_image(:) - mean_face); similarity_score = arrayfun(@(n) 1 / (1 + norm(features(:,n) - feature_vec)), 1:num_images); % find the image with the highest similarity [match_score, match_ix] = max(similarity_score); if match_ix == i cnt = cnt + 1; else end end end result(num_eigenfaces, 1) = cnt; result(num_eigenfaces, 2) = cnt / 120; end figure, plot(result(:, 2)); xlabel('No. of eigenfaces'), ylabel('Correct rate'); xlim([1 39]), ylim([0 1]), grid on;