基于OpenCV一个车牌识别的实例.zip

#include "stdafx.h" #include "LPR.h" bool comp(RotatedRect a, RotatedRect b) { return a.center.x < b.center.x; } float PR(Mat &src, Mat &dst, int num) { int temp = 0; float result = 0; // imshow("src", src); // imshow("dst", dst); // waitKey(0); for (int i = 0; i < src.rows; i++) { for (int j = 0; j < src.cols; j++) { if (src.ptr(i)[j] == dst.ptr(i)[j]) { temp++; } } } result = (float)temp / (src.rows*src.cols); return result; } void Output(int num) { if (num < 10) { cout << num << " "; } else { if (num == 10) { cout << "D" << " "; } else if (num == 11) { cout << "津" << " "; } else if (num == 12) { cout << "H" << " "; } else { cout << "*" << " "; } } } LPR::LPR() { } LPR::LPR(String path) { this->load(path); } void LPR::load(String path) { //读取车牌图片 srcImage = imread(path); if (srcImage.empty()) { cout << "错误的路径!" << endl; /* exit(-1);*/ } cvtColor(srcImage, markedImage, CV_16S); //载入数字图像 for (int i = 0; i < template_len; i++) { stringstream stream; stream << "pictures/num_"; stream << i; stream << ".bmp"; String name = stream.str(); NUM[i] = imread(name); if (NUM[i].empty()) { cout << "未能读取" << name << endl; } threshold(NUM[i], NUM[i], 0, 255, CV_THRESH_BINARY); } } void LPR::showSrc() { //显示原图 imshow("原图",srcImage); } void LPR::gaussFilter() { //高斯滤波 GaussianBlur(srcImage, gaussImage, Size(3, 3), 0); /* imshow("【效果图】高斯滤波", gaussImage);*/ } void LPR::sobel() { //sobel运算 //创建 grad_x 和 grad_y 矩阵 Mat grad_x, grad_y; Mat abs_grad_x, abs_grad_y; //求 X方向梯度 Sobel(gaussImage, grad_x, CV_16S, 1, 0, 3, 1, 1, BORDER_DEFAULT); convertScaleAbs(grad_x, abs_grad_x); /* imshow("【效果图】 X方向Sobel", abs_grad_x);*/ //求Y方向梯度 Sobel(gaussImage, grad_y, CV_16S, 0, 1, 3, 1, 1, BORDER_DEFAULT); convertScaleAbs(grad_y, abs_grad_y); /* imshow("【效果图】Y方向Sobel", abs_grad_y);*/ //合并梯度(近似) addWeighted(abs_grad_x, 0.5, abs_grad_y, 0.5, 0, sobelImage); /* imshow("【效果图】整体方向Sobel", sobelImage);*/ } void LPR::thresholding() { //颜色空间转换函数 cvtColor(sobelImage, binImage, CV_RGB2GRAY); //threshold 方法是通过遍历灰度图中点，将图像信息二值化，处理过后的图片只有二种色值。 threshold(binImage, binImage, 180, 255, CV_THRESH_BINARY); /* imshow("【效果图】二值化处理", binImage);*/ } void LPR::closeOperation() { //定义核 Mat element = getStructuringElement(MORPH_RECT, Size(17, 17)); //进行形态学操作 morphologyEx(binImage, closeImage, MORPH_CLOSE, element); /* imshow("【效果图】闭运算", closeImage);*/ } void LPR::removeLittle() { //腐蚀操作 Mat ele = getStructuringElement(MORPH_RECT, Size(5, 5));//getStructuringElement返回值定义内核矩阵 erode(closeImage, closeImage, ele);//erode函数直接进行腐蚀操作 /* imshow("【效果图】去除小区域", closeImage);*/ //膨胀操作 ele = getStructuringElement(MORPH_RECT, Size(17, 5));//getStructuringElement返回值定义内核矩阵 dilate(closeImage, closeImage, ele);//dilate函数直接进行膨胀操作 /* imshow("【效果图】再膨胀", closeImage);*/ } void LPR::getMaxArea() { //取轮廓 vector< vector< Point> > contours; findContours(closeImage, contours, // a vector of contours CV_RETR_EXTERNAL, // 提取外部轮廓 CV_CHAIN_APPROX_NONE); // all pixels of each contours int max = 0; for (size_t i = 0; i < contours.size(); i++) { vector<Point> p; p = contours[i]; if (p.size() > max) { max = p.size(); maxArea = p; } } //绘制最大区域 for (size_t i = 0; i < maxArea.size(); i++) { circle(markedImage, maxArea[i], 1, Scalar(240, 255, 25)); } //绘制最小外接矩形 minRect = minAreaRect(maxArea); minRect.points(P); for (int j = 0; j <= 3; j++) { line(markedImage, P[j], P[(j + 1) % 4], Scalar(255), 2); } /* imshow("【效果图】标记", markedImage);*/ } void LPR::affine() { //仿射变换 Point2f srcTri[3]; Point2f dstTri[3]; Mat rot_mat(2, 3, CV_32FC1); Mat warp_mat(2, 3, CV_32FC1); //设置三个点来计算仿射变换 //左上 int tep = markedImage.cols; for (size_t i = 0; i < maxArea.size(); i++) { if (maxArea[i].x < tep) { tep = maxArea[i].x; srcTri[0] = maxArea[i]; } } //左下 tep = 0; for (size_t i = 0; i < maxArea.size(); i++) { if (maxArea[i].y > tep) { tep = maxArea[i].y; srcTri[1] = maxArea[i]; } } //右下 tep = 0; for (size_t i = 0; i < maxArea.size(); i++) { if (maxArea[i].x > tep) { tep = maxArea[i].x; srcTri[2] = maxArea[i]; } } //Scalar(blue,green,red) circle(markedImage, srcTri[0], 5, Scalar(255)); circle(markedImage, srcTri[1], 5, Scalar(0, 255)); circle(markedImage, srcTri[2], 5, Scalar(0, 0, 255)); /* imshow("【标记】车牌区域", markedImage);*/ dstTri[0] = Point2f(0, 0); dstTri[1] = Point2f(0, minRect.size.height); dstTri[2] = Point2f(minRect.size.width, minRect.size.height); //计算仿射变换矩阵 warp_mat = getAffineTransform(srcTri, dstTri); //对加载图形进行仿射变换操作 warpAffine(gaussImage, plateImage, warp_mat, minRect.size); /* imshow("【效果图】仿射变换", plateImage);*/ } void LPR::reThreshold() { //再次二值化 threshold(plateImage, reBinImage, 180, 255, CV_THRESH_BINARY_INV); /* imshow("再次二值化", reBinImage);*/ } void LPR::recognition() { cvtColor(reBinImage, reBinImage, COLOR_BGR2GRAY); cvtColor(plateImage, charMarkedImage, CV_16S); threshold(reBinImage, reBinImage, 48, 255, CV_THRESH_BINARY_INV); //运算操作 Mat ele = getStructuringElement(MORPH_RECT, Size(3, 5));//getStructuringElement返回值定义内核矩阵 dilate(reBinImage, reBinImage, ele); /* imshow("膨胀", reBinImage);*/ vector<vector<Point>> contours; vector<Vec4i> hierarchy; findContours(reBinImage, contours, hierarchy, CV_RETR_EXTERNAL, CV_CHAIN_APPROX_NONE); vector<RotatedRect> RectArr(license_len); int a = 0; //画出可包围数字的最小矩形 for (vector<vector<Point>>::iterator It = contours.begin(); It < contours.end(); It++) { Point2f vertex[4]; RotatedRect rect = minAreaRect(*It); if (rect.size.area() > 200) { RectArr[a] = rect; RectArr[a].points(vertex); a++; for (int j = 0; j < 4; j++) line(charMarkedImage, vertex[j], vertex[(j + 1) % 4], Scalar(0, 0, 255), 1); } } imshow("轮廓", charMarkedImage); Mat warp_mat(2, 3, CV_32FC1); sort(RectArr.begin(), RectArr.end(), comp);//对轮廓从左到右排序 cout << "车牌："; for (int i = 0; i < RectArr.size(); i++) { Point2f vertex[4]; RectArr[i].points(vertex); //仿射变换 Point2f srcTri[3]; Point2f dstTri[3]; srcTri[0] = vertex[0]; srcTri[1] = vertex[1]; srcTri[2] = vertex[2]; dstTri[0] = Point2f(0, 400); dstTri[1] = Point2f(0, 0); dstTri[2] = Point2f(400, 0); Mat warp_mat = getAffineTransform(srcTri, dstTri); //对加载图形进行仿射变换操作 warpAffine(plateImage, charMat[i], warp_mat, Size(400, 400)); // stringstream stream;// 字符流 // stream << i + 1; // String name = stream.str(); //二值化 threshold(charMat[i], charMat[i], 150, 255, CV_THRESH_BINARY_INV); float maxRate = 0, tempRate = 0; int mostLike; for (int j = 0; j < template_len; j++) { tempRate = PR(charMat[i], NUM[j], j); /* cout << setprecision(2) << tempRate<<" ";//输出匹配率*/ if (tempRate > maxRate) { maxRate = tempRate; mostLike = j; } } Output(mostLike); /* imwrite(name+".bmp", charMat[i]);//保存提取的字符*/ } cout << endl; } void LPR::processing() { //高斯滤波 gaussFilter(); //sobel运算 sobel(); //二值化 thresholding(); //闭运算 closeOperation(); //去除小区域 removeLittle(); //取最大轮廓 getMaxArea(); //仿射变换 affine(); //再次二值化 reThreshold(); //识别字符 recognition(); } LPR::~LPR() { destroyAllWindows(); }