QT5.9.1+opencv4.1.0车牌识别

#include <windows.h> #include <iostream> #include <opencv2/opencv.hpp> //用于截取图片上的车牌 using namespace cv; using namespace std; using namespace ml; const string chars[34] { "0","1","2","3","4","5","6","7","8","9", "A","B","C","D","E","F","G","H","J","K", "L","M","N","P","Q","R","S","T","U","V", "W","X","Y","Z" }; const string chineses[20] { "浙","皖","京","渝","闽","甘","粤","桂","贵","琼", "冀","黑","豫","鄂","湘","苏","赣","吉","辽","蒙" }; int H_data = 100; int S_data = 125; int V_data = 50; int H_data_max = 125; int S_data_max = 255; int V_data_max = 245; void canny_track1(int, void *) { } void canny_track2(int, void *) { } void canny_track3(int, void *) { } void canny_track4(int, void *) { } void canny_track5(int, void *) { } void canny_track6(int, void *) { } void can_chang() { namedWindow("【HSV色域阈值调节】", WINDOW_NORMAL); cv::createTrackbar("H_MIN", "【HSV色域阈值调节】", &H_data, 255, canny_track1); canny_track1(0, 0); cv::createTrackbar("S_MIN", "【HSV色域阈值调节】", &S_data, 255, canny_track2); canny_track2(0, 0); cv::createTrackbar("V_MIN", "【HSV色域阈值调节】", &V_data, 255, canny_track3); canny_track3(0, 0); cv::createTrackbar("H_MAX", "【HSV色域阈值调节】", &H_data_max, 255, canny_track4); canny_track4(0, 0); cv::createTrackbar("S_MAX", "【HSV色域阈值调节】", &S_data_max, 255, canny_track5); canny_track5(0, 0); cv::createTrackbar("V_MAX", "【HSV色域阈值调节】", &V_data_max, 255, canny_track6); canny_track6(0, 0); } void filter_backdrop(Mat &img) { Mat img_HSV = img.clone();//复制 cvtColor(img, img, COLOR_BGR2GRAY); cvtColor(img_HSV, img_HSV, COLOR_BGR2HSV); int x = 0; int y = 0; for (y = 0; y < img_HSV.rows; y++) { for (x = 0; x < img_HSV.cols; x++) { /*if (x <img_HSV.cols / 4 ||x > img_HSV.cols / 4 * 3) { img.at<uchar>(y, x) = 255; } else */if ((img_HSV.at<Vec3b>(y, x)[0] <= H_data_max&&img_HSV.at<Vec3b>(y, x)[0] >= H_data) && (img_HSV.at<Vec3b>(y, x)[1] <= S_data_max&&img_HSV.at<Vec3b>(y, x)[1] >= S_data) && (img_HSV.at<Vec3b>(y, x)[2] <= V_data_max&&img_HSV.at<Vec3b>(y, x)[2] >= V_data)) { img.at<uchar>(y, x) = 255; } else { img.at<uchar>(y, x) = 0; } } } //imshow("【滤除背景后的灰度图】", img); } const int image_cols = 16; const int image_rows = 20; void discern_car(Mat image,vector<Mat> & img_buff,string &chr_chars)//识别图片 { Ptr<ANN_MLP> bp_chinese = StatModel::load<ANN_MLP>("../bpcharModel_chinese.xml"); Ptr<ANN_MLP> bp_char = StatModel::load<ANN_MLP>("../bpcharModel_16_20.xml"); Mat grayImg = image.clone(); filter_backdrop(grayImg); //Candy/sobel 边缘检测： Mat candy_image; Canny(grayImg, candy_image, 500, 200, 3); //imshow("边沿检测", candy_image); //形态学处理 //图片膨胀处理 Mat dilate_image, erode_image; //自定义 核进行 x 方向的膨胀腐蚀 Mat elementX = getStructuringElement(MORPH_RECT, Size(40, 1)); Mat elementY = getStructuringElement(MORPH_RECT, Size(1, 19)); Point point(-1, -1); dilate(candy_image, dilate_image, elementX, point, 2); erode(dilate_image, erode_image, elementX, point, 4); dilate(erode_image, dilate_image, elementX, point, 2); //自定义 核进行 Y 方向的膨胀腐蚀 erode(dilate_image, erode_image, elementY, point, 1); dilate(erode_image, dilate_image, elementY, point, 2); //噪声处理 //平滑处理 中值滤波 Mat blurr_image; medianBlur(dilate_image, blurr_image, 3); //imshow("腐蚀膨胀", blurr_image); //矩形轮廓查找与筛选： Mat contour_image; //深拷贝 查找轮廓会改变源图像信息，需要重新 拷贝 图像 contour_image = blurr_image.clone(); //查找图像轮廓 Mat midImg3(contour_image.rows, contour_image.cols, CV_8UC3, Scalar::all(0)); vector<vector<Point>> contours; findContours(contour_image, contours, RETR_CCOMP, CHAIN_APPROX_SIMPLE); //轮廓表示为一个矩形 车牌提取 Mat roi_image; //截取车牌图像 for (int i = 0; i < contours.size(); i++) { Rect r = boundingRect(Mat(contours[i])); //RotatedRect r = minAreaRect(Mat(contours[i])); if ((float)r.width / r.height >= 2.2 && (float)r.width / r.height <= 3.6) { rectangle(contour_image, r, Scalar(0, 0, 255), 2); roi_image = image(r); } } //重映射大小 resize(roi_image, roi_image, Size(210, 70), (0, 0), (0, 0), INTER_AREA);//使用象素关系重采样。当图像缩小时候，该方法可以避免波纹出现 img_buff.push_back(roi_image);//车牌图片 //灰度图 Mat roi_gray_image; cvtColor(roi_image, roi_gray_image, COLOR_BGR2GRAY); //腐蚀膨胀消除黑点 Mat src = roi_gray_image.clone();//复制 Mat dst; Mat structureElement = getStructuringElement(MORPH_RECT, Size(4, 3), Point(-1, -1)); //创建结构元 dilate(src, dst, structureElement, Point(-1, -1), 1); //调用膨胀API erode(dst, dst, structureElement); //调用腐蚀API Mat test_image;//复制 threshold(dst, test_image, 0, 255,THRESH_OTSU); Mat gray = test_image.clone(); Mat grays = test_image.clone(); vector<vector<Point>> contours_buff; //定义轮廓集合 vector<Vec4i> hierarchy; findContours(gray, contours_buff, hierarchy, RETR_TREE, CHAIN_APPROX_NONE);//CV_RETR_EXTERNAL只检测外部轮廓 vector<Rect> Rects;//用于保存边沿框 int width_v = 0; for (int index_ori = 0; index_ori >= 0; index_ori = hierarchy[index_ori][0]) { if (contours_buff.size() == 0)break; if (index_ori < 0)break; drawContours(grays, contours_buff, index_ori, Scalar(255), 1, 8, hierarchy); Rect rect = boundingRect(contours_buff[index_ori]);//检测外轮廓 rect.x-=1; rect.y-=1; rect.width += 2; rect.height += 2; if (rect.x < 0)rect.x = 0; if (rect.y < 0)rect.y = 0; if (rect.width + rect.x > roi_gray_image.cols)rect.width = roi_gray_image.cols - rect.x; if (rect.height + rect.y > roi_gray_image.rows)rect.height = roi_gray_image.rows - rect.y; if (rect.width<roi_gray_image.cols /7 && rect.height>roi_gray_image.rows *0.6 && rect.height<roi_gray_image.rows *0.85)//一个车牌最多7个字符 进行初步的筛选 { Rects.push_back(rect);//记录后对容器内的数据进行排序 基本就是需要的边沿 if ((float)rect.width / rect.height >= 0.3 && (float)rect.width / rect.height <= 0.7)//锁定长宽比例 //这个基本是正确的 { //计算平均长度 width_v += rect.width; width_v /= 2; } } } //对边沿进行排序和适当处理 for (int i=0; i < Rects.size(); i++) { if (!((float)Rects[i].width / Rects[i].height >= 0.3 && (float)Rects[i].width / Rects[i].height <= 0.7))//锁定长宽比例 //这个基本是正确的 { Rects[i].x =Rects[i].x +(Rects[i].width - width_v) / 2; Rects[i].width = width_v; if (Rects[i].width + Rects[i].x > roi_gray_image.cols)Rects[i].width = roi_gray_image.cols - Rects[i].x; if (Rects[i].height + Rects[i].y > roi_gray_image.rows)Rects[i].height = roi_gray_image.rows - Rects[i].y; } } //排序 for (int i = Rects.size(); i >=0; i--) { for (int x = 0; x < i-1; x++) { if (Rects[x + 1].x < Rects[x].x) { Rect Rects_buff; Rects_buff.x = Rects[x].x; Rects_buff.y = Rects[x].y; Rects_buff.width = Rects[x].width; Rects_buff.height= Rects[x].height; Rects[x].x = Rects[x+1].x; Rects[x].y = Rects[x+1].y; Rects[x].width = Rects[x+1].width; Rects[x].height = Rects[x+1].height; Rects[x+1].x = Rects_buff.x