您好,登錄后才能下訂單哦!
這篇文章給大家介紹使用opencv怎么實現一個車道線檢測功能,內容非常詳細,感興趣的小伙伴們可以參考借鑒,希望對大家能有所幫助。
原理:
算法基本思想說明:
傳統的車道線檢測,多數是基于霍夫直線檢測,其實這個里面有個很大的誤區,霍夫直線擬合容易受到各種噪聲干擾,直接運用有時候效果不好,更多的時候通過霍夫直線檢測進行初步的篩選,然后再有針對性的進行直線擬合,根據擬合的直線四個點坐標,繪制出車道線,這種方式可以有效避免霍夫直線擬合不良后果,是一種更加穩定的車道線檢測方法,在實際項目中,可以選擇兩種方法并行,在計算出結果后進行疊加或者對比提取,今天分享的案例主要是繞開了霍夫直線檢測,通過對二值圖像進行輪廓分析與幾何分析,提取到相關的車道線信息、然后進行特定區域的像素掃描,擬合生成直線方程,確定四個點繪制出車道線,對連續的視頻來說,如果某一幀無法正常檢測,就可以通過緩存來替代繪制,從而實現在視頻車道線檢測中實時可靠。
原理圖:
代碼:
#include <opencv2/opencv.hpp> #include <iostream> #include <cmath> using namespace cv; using namespace std; /** **1、讀取視頻 **2、二值化 **3、輪廓發現 **4、輪廓分析、面積就算,角度分析 **5、直線擬合 **6、畫出直線 ** */ Point left_line[2]; Point right_line[2]; void process(Mat &frame, Point *left_line, Point *right_line); Mat fitLines(Mat &image, Point *left_line, Point *right_line); int main(int argc, char** argv) { //讀取視頻 VideoCapture capture("E:/opencv/road_line.mp4"); int height = capture.get(CAP_PROP_FRAME_HEIGHT); int width = capture.get(CAP_PROP_FRAME_WIDTH); int count = capture.get(CAP_PROP_FRAME_COUNT); int fps = capture.get(CAP_PROP_FPS); //初始化 left_line[0] = Point(0,0); left_line[1] = Point(0, 0); right_line[0] = Point(0, 0); right_line[1] = Point(0, 0); cout << height<<" "<< width<< " " <<count<< " " <<fps << endl; //循環讀取視頻 Mat frame; while (true) { int ret = capture.read(frame); if (!ret) { break; } imshow("input", frame); process(frame, left_line, right_line); char c = waitKey(5); if (c == 27) { break; } } } void process(Mat &frame, Point *left_line, Point *right_line ){ Mat gray,binary; /**灰度化*/ cvtColor(frame, gray, COLOR_BGR2GRAY); //threshold(gray, binary, ); //邊緣檢測 Canny(gray, binary, 150, 300); //imshow("Canny", binary); for (size_t i = 0; i < (gray.rows/2+40); i++) { for (size_t j = 0; j < gray.cols; j++) { binary.at<uchar>(i, j) = 0; } } imshow("binary", binary); //尋找輪廓 vector<vector<Point>> contours; findContours(binary, contours, RETR_EXTERNAL, CHAIN_APPROX_SIMPLE); Mat out_image = Mat::zeros(gray.size(), gray.type()); for (int i = 0; i < contours.size(); i++) { //計算面積與周長 double length = arcLength(contours[i], true); double area = contourArea(contours[i]); //cout << "周長 length:" << length << endl; //cout << "面積 area:" << area << endl; //外部矩形邊界 Rect rect = boundingRect(contours[i]); int h = gray.rows - 50; //輪廓分析: if (length < 5.0 || area < 10.0) { continue; } if (rect.y > h) { continue; } //最小包圍矩形 RotatedRect rrt = minAreaRect(contours[i]); //cout << "最小包圍矩形 angle:" << rrt.angle << endl; double angle = abs(rrt.angle); //angle < 50.0 || angle>89.0 if (angle < 20.0 || angle>84.0) { continue; } if (contours[i].size() > 5) { //用橢圓擬合 RotatedRect errt = fitEllipse(contours[i]); //cout << "用橢圓擬合err.angle:" << errt.angle << endl; if ((errt.angle<5.0) || (errt.angle>160.0)) { if (80.0 < errt.angle && errt.angle < 100.0) { continue; } } } //cout << "開始繪制:" << endl; drawContours(out_image, contours, i, Scalar(255), 2, 8); imshow("out_image", out_image); } Mat result = fitLines(out_image, left_line, right_line); imshow("result", result); Mat dst; addWeighted(frame, 0.8, result, 0.5,0, dst); imshow("lane-lines", dst); } //直線擬合 Mat fitLines(Mat &image, Point *left_line, Point *right_line) { int height = image.rows; int width = image.cols; Mat out = Mat::zeros(image.size(), CV_8UC3); int cx = width / 2; int cy = height / 2; vector<Point> left_pts; vector<Point> right_pts; Vec4f left; for (int i = 100; i < (cx-10); i++) { for (int j = cy; j < height; j++) { int pv = image.at<uchar>(j, i); if (pv == 255) { left_pts.push_back(Point(i, j)); } } } for (int i = cx; i < (width-20); i++) { for (int j = cy; j < height; j++) { int pv = image.at<uchar>(j, i); if (pv == 255) { right_pts.push_back(Point(i, j)); } } } if (left_pts.size() > 2) { fitLine(left_pts, left, DIST_L1, 0, 0.01, 0.01); double k1 = left[1] / left[0]; double step = left[3] - k1 * left[2]; int x1 = int((height - step) / k1); int y2 = int((cx - 25)*k1 + step); Point left_spot_1 = Point(x1, height); Point left_spot_end = Point((cx - 25), y2); line(out, left_spot_1, left_spot_end, Scalar(0, 0, 255), 8, 8, 0); left_line[0] = left_spot_1; left_line[1] = left_spot_end; } else { line(out, left_line[0], left_line[1], Scalar(0, 0, 255), 8, 8, 0); } if (right_pts.size()>2) { Point spot_1 = right_pts[0]; Point spot_end = right_pts[right_pts.size()-1]; int x1 = spot_1.x; int y1 = spot_1.y; int x2 = spot_end.x; int y2 = spot_end.y; line(out, spot_1, spot_end, Scalar(0, 0, 255), 8, 8, 0); right_line[0] = spot_1; right_line[1] = spot_end; } else { line(out, right_line[0], right_line[1], Scalar(0, 0, 255), 8, 8, 0); } return out; }
關于使用opencv怎么實現一個車道線檢測功能就分享到這里了,希望以上內容可以對大家有一定的幫助,可以學到更多知識。如果覺得文章不錯,可以把它分享出去讓更多的人看到。
免責聲明:本站發布的內容(圖片、視頻和文字)以原創、轉載和分享為主,文章觀點不代表本網站立場,如果涉及侵權請聯系站長郵箱:is@yisu.com進行舉報,并提供相關證據,一經查實,將立刻刪除涉嫌侵權內容。