MATLAB神经网络原理与实例精解_matlab_基于概率神经网络的手写体数字识别

function [Feature,bmp,flag]=getFeature1(A) % getFeature.m % 提取64*64二值图像的特征向量 % input: % A: 64*64矩阵 % output: % Feature: 长度为14的特征向量 % bmp : 图像中的数字部分 % flag : 标志位，表示数字部分的宽高比 % 反色 A = ones(64) - A; % 提取数字部分 [x, y] = find(A == 1); % 截取图像中的数字部分 A = A(min(x):max(x),min(y):max(y)); % 计算宽高比和标志位 flag = (max(y)-min(y)+1)/(max(x)-min(x)+1); if flag < 0.5 flag = 0; elseif flag >=0.5 && flag <0.75 flag = 1; elseif flag >=0.75 && flag <1 flag = 2; else flag = 3; end % 重新放大，将长或宽调整为64 rate = 64 / max(size(A)); % 调整尺寸 A = imresize(A,rate); [x,y] = size(A); % 不足64的部分用零填充 if x ~= 64 A = [zeros(ceil((64-x)/2)-1,y);A;zeros(floor((64-x)/2)+1,y)]; end; if y ~= 64 A = [zeros(64,ceil((64-y)/2)-1),A,zeros(64,floor((64-y)/2)+1)]; end %% 三条竖线与数字字符的交点个数 F(1)~F(3) % 1/2 竖线交点数量 Vc = 32; Num = 0; for i = 1:64 Num = Num+A(i, Vc); end F(1) = Num; % F(1) = sum(A(:,Vc)); % 5/12 竖线交点数量 Vc = round(64*3/12); Num = 0; for i = 1:64 Num = Num + A(i, Vc); end F(2) = Num; % F(2) = sum(A(:,Vc)); % 7/12 竖线交点数量 Vc = round(64*9/12); Num = 0; for i = 1:64 Num = Num + A(i, Vc); end F(3)=Num; % F(3) = sum(A(:,Vc)); %% 三条横线与数字字符的交点个数 F(4)~F(6) % 1/2 水平线交点数量 Hc = 32; Num = 0; for i = 1:64 Num = Num + A(Hc, i); end F(4) = Num; % F(4) = sum(A(Hc,:)); % 1/3 水平线处交点数量， Hc = round(64/3); Num = 0; for i = 1:64 Num = Num + A(Hc, i); end F(5) = Num; % F(5) = sum(A(Hc,:)); % 2/3水平线处交点数量 Hc = round(2*64/3); Num = 0; for i = 1:64 Num = Num + A(Hc, i); end F(6) = Num; % F(6) = sum(A(Hc,:)); %% 两条对角线的交点数量 % 主对角线交点数， x3 = 1; y3 = 1; Num = 0; for i = 0:63 Num = Num+A(x3+i,y3+i); end F(7) = Num; % F(7) = sum(diag(A)); % 次对角线交点数 x4 = 1; y4 = 64; Num = 0; for i = 0:63 Num = Num + A(x4+i, y4-i); end F(8) = Num; % F(8) = sum(diag(rot90(A))); %% 小方块 % 右下角1/2小方块中的所有点 Num = 0; for i = 32:64 for r = 32:64 Num = Num + A(i,r); end end F(9) = Num/10; % t = A(32:64,33:64); % F(9) = sum(t(:))/10; % 左上角1/2小方块中的所有点 Num = 0; for i3 = 1:32 for r3 = 1:32 Num = Num + A(i3,r3); end end F(10) = Num/10; % t = A(1:32,1:32); % F(10) = sum(t(:))/10; % 左下角方块中的所有点 Num = 0; for i4 = 1:32 for r4 = 32:64 Num = Num + A(i4,r4); end end F(11) = Num/10; % t = A(1:32,32:64); % F(11) = sum(t(:))/10; % 右上角方块中的所有点 Num = 0; for i5 = 32:64 for r5 = 1:32 Num = Num + A(i5,r5); end end F(12) = Num/10; % t = A(32:64,1:32); % F(12) = sum(t(:))/10; % 下方2/3部分的所有像素点 Num = 0; for i1 = 1:64 for r1 = 16:48 Num = Num + A(i1,r1); end end F(13) = Num/20; % t = A(1:64,16:64); % F(13) = sum(t(:))/20; % 右方2/3部分的所有像素点 Num = 0; for i2 = 16:48 for r2 = 1:64 Num = Num + A(i2,r2); end end F(14) = Num/20; % t = A(16:48,1:64); % F(14) = sum(t(:))/20; Feature = F'; bmp = A;