行人再识别代码，基于巴氏距离用matlab实现，程序简单，容易理解，

clear all; close all; %从文件夹中读取图片.a查询库，b候选行人库 img_dir1 = dir('./cam_a/*.bmp'); img_dir2 = dir('./cam_b/*.bmp'); %统计两个文件夹中的图片个数 num1=length(img_dir1); num2=length(img_dir2); S=rand(num1,num2); ji=rand(1,50); %k就是最后排序的前k个 image = imread(['./cam_a/' img_dir1(1).name]); [x,y,z]=size(image); %x:高 y：宽 z:维度 for i = 1:length(img_dir1) I = imread(['./cam_a/' img_dir1(i).name]); %将I的躯干和腿部分块(躯干从0.2-0.6，腿从0.6-1，分为5块) I1=I(0.2*x:0.28*x,0.1*y:0.9*y,1:3); %选取目标块，Q=I（x1:x2,y1:y2.z1:z2）;z是维度1:3表示彩色。二值图不需要z I2=I(0.28*x:0.36*x,0.1*y:0.9*y,1:3); I3=I(0.36*x:0.44*x,0.1*y:0.9*y,1:3); I4=I(0.44*x:0.52*x,0.1*y:0.9*y,1:3); I5=I(0.52*x:0.60*x,0.1*y:0.9*y,1:3); I6=I(0.60*x:0.68*x,0.1*y:0.9*y,1:3); I7=I(0.68*x:0.76*x,0.1*y:0.9*y,1:3); I8=I(0.76*x:0.84*x,0.1*y:0.9*y,1:3); I9=I(0.84*x:0.92*x,0.1*y:0.9*y,1:3); I10=I(0.92*x:1*x,0.1*y:0.9*y,1:3); for j=1:length(img_dir2) %实现了文件夹里的文件读取（好像要求读取格式一） W= imread(['./cam_b/' img_dir2(j).name]); % %尺度变换为175*70 % I=imresize(I,[175,70]); % W=imresize(W,[175,70]); %对W分块 W1=W(0.2*x:0.28*x,0.1*y:0.9*y,1:3); %选取目标块，Q=I（x1:x2,y1:y2.z1:z2）;z是维度1:3表示彩色。二值图不需要z W2=W(0.28*x:0.36*x,0.1*y:0.9*y,1:3); W3=W(0.36*x:0.44*x,0.1*y:0.9*y,1:3); W4=W(0.44*x:0.52*x,0.1*y:0.9*y,1:3); W5=W(0.52*x:0.60*x,0.1*y:0.9*y,1:3); W6=W(0.60*x:0.68*x,0.1*y:0.9*y,1:3); W7=W(0.68*x:0.76*x,0.1*y:0.9*y,1:3); W8=W(0.76*x:0.84*x,0.1*y:0.9*y,1:3); W9=W(0.84*x:0.92*x,0.1*y:0.9*y,1:3); W10=W(0.92*x:1*x,0.1*y:0.9*y,1:3); %计算每一块的相似度 [S1] = similarity( I1,W1 ); [S2] = similarity( I2,W2 ); [S3] = similarity( I3,W3 ); [S4] = similarity( I4,W4 ); [S5] = similarity( I5,W5 ); [S6] = similarity( I6,W6 ); [S7] = similarity( I7,W7 ); [S8] = similarity( I8,W8 ); [S9] = similarity( I9,W9 ); [S10] = similarity( I10,W10 ); %计算平均相似度 S(i,j)=(S1+S2+S3+S4+S5+S6+S7+S8+S9+S10)/10; end end for i=1:num1 a=S(i,:); [b,ind]=sort(a); for j=1:50 if ind(j)==i ji(i)=j; break end end end %非常牛逼的一个函数，可以统计ji中每个元素出现的个数以及所占的比例，cmc曲线问题可以直接解决。 rank=tabulate(ji); % f=f'; % c=f(1,:); %取f中第一行元素 % d=f(3,:); %取f中第三行元素 % for j=2:k % d(j)=d(j)+d(j-1); %求cmc曲线的纵坐标 % end % % 画出cmc曲线 % plot(c,d); % title('CMC曲线'); % xlabel('rank'); % ylabel('匹配率'); % %保存 % save VIPeRtest; % for m=1:3 % W= imread(['./test/' img_dir2(ind(m)).name]); % figure(); % imshow(W); % end % I=imread('S01202000.png'); % I=imresize(I,[175,70]); % W=imread('S01203000.png'); % W=imresize(W,[175,70]); % [x,y,z]=size(I); %x:高 y：宽 z:维度 % % % 将I的躯干和腿部分块(躯干从0.2-0.6，腿从0.6-1，分为5块) % I1=I(0.2*x:0.28*x,0.1*y:0.9*y,1:3); %选取目标块，Q=I（x1:x2,y1:y2.z1:z2）;z是维度1:3表示彩色。二值图不需要z % I2=I(0.28*x:0.36*x,0.1*y:0.9*y,1:3); % I3=I(0.36*x:0.44*x,0.1*y:0.9*y,1:3); % I4=I(0.44*x:0.52*x,0.1*y:0.9*y,1:3); % I5=I(0.52*x:0.60*x,0.1*y:0.9*y,1:3); % I6=I(0.60*x:0.68*x,0.1*y:0.9*y,1:3); % I7=I(0.68*x:0.76*x,0.1*y:0.9*y,1:3); % I8=I(0.76*x:0.84*x,0.1*y:0.9*y,1:3); % I9=I(0.84*x:0.92*x,0.1*y:0.9*y,1:3); % I10=I(0.92*x:1*x,0.1*y:0.9*y,1:3); % % 对W分块 % W1=W(0.2*x:0.28*x,0.1*y:0.9*y,1:3); %选取目标块，Q=I（x1:x2,y1:y2.z1:z2）;z是维度1:3表示彩色。二值图不需要z % W2=W(0.28*x:0.36*x,0.1*y:0.9*y,1:3); % W3=W(0.36*x:0.44*x,0.1*y:0.9*y,1:3); % W4=W(0.44*x:0.52*x,0.1*y:0.9*y,1:3); % W5=W(0.52*x:0.60*x,0.1*y:0.9*y,1:3); % W6=W(0.60*x:0.68*x,0.1*y:0.9*y,1:3); % W7=W(0.68*x:0.76*x,0.1*y:0.9*y,1:3); % W8=W(0.76*x:0.84*x,0.1*y:0.9*y,1:3); % W9=W(0.84*x:0.92*x,0.1*y:0.9*y,1:3); % W10=W(0.92*x:1*x,0.1*y:0.9*y,1:3); % % 计算每一块的相似度 % [S1] = similarity( I1,W1 ); % [S2] = similarity( I2,W2 ); % [S3] = similarity( I3,W3 ); % [S4] = similarity( I4,W4 ); % [S5] = similarity( I5,W5 ); % [S6] = similarity( I6,W6 ); % [S7] = similarity( I7,W7 ); % [S8] = similarity( I8,W8 ); % [S9] = similarity( I9,W9 ); % [S10] = similarity( I10,W10 ); % % 计算平均相似度 % S=(S1+S2+S3+S4+S5+S6+S7+S8+S9+S10)/10