验证码识别技术

此验证码识别技术有预处理、特征提取、神经网络识别几部分，参考过车牌识别技术。程序实现分样本下载器、用户交互程序、MLP建构程序三块，使用OpenCV，喜欢的可以留意下AnnMLP的使用方法，国内这方面的资料还是比较少的，欢迎交流QQ:13895354。

 

预处理技术是专门针对某种特定验证码设计的，其特点有：英文+数字，彩色，倾斜，无粘连，有不规则噪声，有删除线，至于具体哪家网站的验证码，我就不方便透露了

（用户交互程序）

#include <cv.h> #include <highgui.h> #include <ml.h> #include <iostream> using namespace std; int main( int argc, char** argv ) { cvNamedWindow("1"); cvNamedWindow("2"); IplConvKernel* se = cvCreateStructuringElementEx(2, 2, 1, 1, CV_SHAPE_CROSS); CvANN_MLP mlp; mlp.load( "mpl.xml" ); for(int i=1; i<=100; i++) { // 产生文件名 char fn[1024]; sprintf(fn, ".//sample//%d.jpg", i); // 读取图片 IplImage* gray = cvLoadImage(fn, CV_LOAD_IMAGE_GRAYSCALE); // 二值化 cvThreshold(gray, gray, 180, 255, CV_THRESH_BINARY | CV_THRESH_OTSU); // 去边框 cvRectangle(gray, cvPoint(0, 0), cvPoint(gray->width-1, gray->height-1), CV_RGB(255, 255, 255)); // 去噪 cvDilate(gray, gray, se); /* // 调整角度 cvShowImage("1", gray); IplImage* rote = cvCreateImage( cvGetSize(gray), IPL_DEPTH_8U, 1 ); double t = tan(10.0 / 180.0 * CV_PI); int w = gray->width; int h = gray->height; for(int i = 0; i<h; i++) { unsigned char* lineGray = (unsigned char*)gray->imageData + gray->widthStep * i; unsigned char* lineRote = (unsigned char*)rote->imageData + rote->widthStep * i; for(int j = 0; j<w; j++) { int j2 = j - ((int)(i*t+0.5)); if (j2<0) j2+=w; *(lineRote+j) = *(lineGray+j2); } } cvCopy(rote, gray); cvReleaseImage(&rote); */ // 计算连通域contour cvXorS(gray, cvScalarAll(255), gray, 0); cvShowImage("1", gray); CvMemStorage* storage = cvCreateMemStorage(); CvSeq* contour = NULL; cvFindContours(gray, storage, &contour, sizeof(CvContour), CV_RETR_CCOMP, CV_CHAIN_APPROX_SIMPLE); // 分析连通域 CvSeq* p = contour; while(p) { CvRect rect = cvBoundingRect(p, 0); if( rect.height>=10 )// 文字需要有10像素高度 { // 绘制该连通区域到character cvZero(gray); IplImage* character = cvCreateImage(cvSize(rect.width, rect.height), IPL_DEPTH_8U, 1); cvZero(character); cvDrawContours(character, p, CV_RGB(255, 255, 255), CV_RGB(0, 0, 0), -1, -1, 8, cvPoint(-rect.x, -rect.y)); // 归一化 IplImage* normal = cvCreateImage(cvSize(16, 16), IPL_DEPTH_8U, 1); cvResize(character, normal, CV_INTER_AREA); cvThreshold(normal, normal, 0, 255, CV_THRESH_BINARY | CV_THRESH_OTSU);// 修正 // 计算输入向量 float input[256]; for(int i=0; i<256; i++) input[i] = (normal->imageData[i]==-1); #if 0 // 用户输入结果 cvShowImage("2", normal); char c = cvWaitKey(0); if(c==27) return 0; // 编码0-9:0-9 a-z:10-35 unsigned char cc = 255; if(c>='A'&&c<='Z') cc=c-'A'+10; else if(c>='a'&&c<='z') cc=c-'a'+10; else if(c>='0'&&c<='9') cc=c-'0'; if(cc!=255) { // 转换成输出向量 float output[36]; for(int i=0; i<36; i++) output[i] = 0.0f; output[cc] = 1.0f; // 存储到批处理文件 static FILE* batch = fopen("batch", "wb"); fwrite(output, 4*36, 1, batch); fwrite(input, 4*256, 1, batch); static int count = 0; cout<<count++<<endl; } #else // 识别 CvMat* output = cvCreateMat( 1, 36, CV_32F ); CvMat inputMat = cvMat( 1, 256, CV_32F, input); mlp.predict( &inputMat, output ); CvPoint max_loc = {0,0}; cvMinMaxLoc( output, NULL, NULL, NULL, &max_loc, NULL ); int best = max_loc.x;// 识别结果 char c = (char)( best<10 ? '0'+best : 'A'+best-10 ); cout<<c<<"("<<rect.x<<")"<<" "; cvReleaseMat( &output ); #endif cvReleaseImage(&character); cvReleaseImage(&normal); } p = p->h_next; } cout<<endl; cvWaitKey(0); cvReleaseMemStorage(&storage); cvReleaseImage(&gray); } cvReleaseStructuringElement(&se); cvDestroyAllWindows(); return 0; }

 

事实上更有价值的是OpenCV中的AnnMLP，其技术很成熟，参考了letter_recg范例程序

 （MLP建构程序）

#include <iostream> #include <ml.h> using namespace std; void print_mat(CvMat& mat) { int count = 0; for(int i=0; i<4/*mat.rows*/; i++) { for(int j=0; j<mat.cols; j++) { cout<<mat.data.fl[i*(mat.step/4)+j]/*<<" "*/; } cout<<endl<<endl; } } int main( int argc, char *argv[] ) { // 读入结果responses 特征data FILE* f = fopen( "batch", "rb" ); fseek(f, 0l, SEEK_END); long size = ftell(f); fseek(f, 0l, SEEK_SET); int count = size/4/(36+256); CvMat* batch = cvCreateMat( count, 36+256, CV_32F ); fread(batch->data.fl, size-1, 1, f); CvMat outputs, inputs; cvGetCols(batch, &outputs, 0, 36); cvGetCols(batch, &inputs, 36, 36+256); // 新建MPL CvANN_MLP mlp; int layer_sz[] = { 256, 20, 36 }; CvMat layer_sizes = cvMat( 1, 3, CV_32S, layer_sz ); mlp.create( &layer_sizes ); // 训练 system( "time" ); mlp.train( &inputs, &outputs, NULL, NULL, CvANN_MLP_TrainParams(cvTermCriteria(CV_TERMCRIT_ITER,300,0.01), CvANN_MLP_TrainParams::RPROP, 0.01) ); system( "time" ); // 存储MPL mlp.save( "mpl.xml" ); // 测试 int right = 0; CvMat* output = cvCreateMat( 1, 36, CV_32F ); for(int i=0; i<count; i++) { CvMat input; cvGetRow( &inputs, &input, i ); mlp.predict( &input, output ); CvPoint max_loc = {0,0}; cvMinMaxLoc( output, NULL, NULL, NULL, &max_loc, NULL ); int best = max_loc.x;// 识别结果 int ans = -1;// 实际结果 for(int j=0; j<36; j++) { if( outputs.data.fl[i*(outputs.step/4)+j] == 1.0f ) { ans = j; break; } } cout<<(char)( best<10 ? '0'+best : 'A'+best-10 ); cout<<(char)( ans<10 ? '0'+ans : 'A'+ans-10 ); if( best==ans ) { cout<<"+"; right++; } //cin.get(); cout<<endl; } cvReleaseMat( &output ); cout<<endl<<right<<"/"<<count<<endl; cvReleaseMat( &batch ); system( "pause" ); return 0; }

 

上面的代码都是vc6控制台程序，OpenCV入门的都懂的吧。