character_recognition-master.rar_字符识别_神经字符识别_神经网络_神经网络字符

clear; clc; %16*16的五幅图片 I1=imread('1.bmp'); yb1=im2bw(I1); I2=imread('2.bmp'); yb2=im2bw(I2); I3=imread('3.bmp'); yb3=im2bw(I3); I4=imread('4.bmp'); yb4=im2bw(I4); I5=imread('5.bmp'); yb5=im2bw(I5); p0=zeros(256,5); %转化格式用以生成训练样本 T=zeros(256,5); ta=eye(5); yb=[yb1,yb2,yb3,yb4,yb5]; for k=1:5 for m=1:16 for n=1:16 T(n+16*(m-1),k)=yb(m,n+16*(k-1)); end end end [R1,Q1]=size(T); [R2,Q2]=size(ta); %给理想样本加噪 J=imnoise(T,'gaussian',0.01,0.001); pause %敲任意键开始生成BP网络 clc %生成神经网络 S1=10; S2=R2; net=newff(minmax(T),[S1,S2],{'logsig' 'logsig'},'traingdx'); net.LW{2,1}=net.LW{2,1}*0.01; net.b{2}=net.b{2}*0.01; pause %敲任意键开始用理想样本训练网络 clc %用理想样本训练神经网络 net.performFcn='sse'; net.trainParam.goal=0.01; %训练误差 net.trainParam.epochs=1000; %训练步数 net.trainParam.mc=0.02; %效率 [net,tr]=train(net,T,ta); pause clc %含噪声的“三”作为测试对象 ya3=double(yb3); t3=imnoise(ya3,'gaussian',0,0.01); imshow(t3);title('gaussian 3'); pause clc %对输入模式进行识别 a3=zeros(256,1); for m=1:16 for n=1:16 a3(n+16*(m-1),1)=yb3(m,n); end end result=sim(net,a3); [resultmax,r]=max(result) %取最大者为一的位置作为识别结果 re=eye(16);%绘制识别结果位图 for m=1:16 for n=1:16 re(m,n)=T(n+16*(m-1),r); end end imshow(re);title('result');