基于c++ opencv的车道线检测算法流程及完整代码

以下是基于C++和OpenCV的车道线检测算法流程及完整代码：

1. 读取图像并进行灰度化和高斯滤波

Mat image = imread("image.jpg");
Mat grayImage, blurImage;
cvtColor(image, grayImage, COLOR_BGR2GRAY);
GaussianBlur(grayImage, blurImage, Size(5, 5), 0, 0);

2. 进行Canny边缘检测

Mat edges;
Canny(blurImage, edges, 50, 150);

3. 对图像进行ROI裁剪

Mat roiMask = Mat::zeros(image.size(), image.type());
Point points[4] = {Point(0, image.rows), Point(image.cols / 2, image.rows / 2), Point(image.cols / 2 + 50, image.rows / 2), Point(image.cols, image.rows)};
fillConvexPoly(roiMask, points, 4, Scalar(255, 0, 0));
Mat roiImage;
bitwise_and(edges, roiMask, roiImage);

4. 进行霍夫变换检测直线

vector<Vec2f> lines;
HoughLines(roiImage, lines, 1, CV_PI / 180, 100, 0, 0);

5. 进行直线筛选和拟合

vector<Vec2f> leftLines, rightLines;
for (int i = 0; i < lines.size(); i++) {
    float rho = lines[i][0], theta = lines[i][1];
    if (theta < CV_PI / 4 || theta > 3 * CV_PI / 4) {
        leftLines.push_back(lines[i]);
    } else {
        rightLines.push_back(lines[i]);
    }
}
Vec2f leftLine, rightLine;
if (leftLines.size() > 0) {
    float leftRho = 0, leftTheta = 0;
    for (int i = 0; i < leftLines.size(); i++) {
        leftRho += leftLines[i][0];
        leftTheta += leftLines[i][1];
    }
    leftRho /= leftLines.size();
    leftTheta /= leftLines.size();
    leftLine = Vec2f(leftRho, leftTheta);
}
if (rightLines.size() > 0) {
    float rightRho = 0, rightTheta = 0;
    for (int i = 0; i < rightLines.size(); i++) {
        rightRho += rightLines[i][0];
        rightTheta += rightLines[i][1];
    }
    rightRho /= rightLines.size();
    rightTheta /= rightLines.size();
    rightLine = Vec2f(rightRho, rightTheta);
}

6. 计算直线与图像边界的交点

Point leftStart, leftEnd, rightStart, rightEnd;
if (leftLine[1] != 0) {
    float leftSlope = -1 / tan(leftLine[1]);
    float leftIntercept = leftLine[0] / sin(leftLine[1]);
    leftStart = Point(0, leftIntercept);
    leftEnd = Point(image.cols, leftSlope * image.cols + leftIntercept);
} else {
    leftStart = Point(leftLine[0], 0);
    leftEnd = Point(leftLine[0], image.rows);
}
if (rightLine[1] != 0) {
    float rightSlope = -1 / tan(rightLine[1]);
    float rightIntercept = rightLine[0] / sin(rightLine[1]);
    rightStart = Point(0, rightIntercept);
    rightEnd = Point(image.cols, rightSlope * image.cols + rightIntercept);
} else {
    rightStart = Point(rightLine[0], 0);
    rightEnd = Point(rightLine[0], image.rows);
}

7. 绘制车道线并显示图像

line(image, leftStart, leftEnd, Scalar(0, 0, 255), 2);
line(image, rightStart, rightEnd, Scalar(0, 0, 255), 2);
imshow("Lane Detection", image);

完整代码如下：

#include <opencv2/opencv.hpp>

using namespace cv;

int main() {
    Mat image = imread("image.jpg");
    Mat grayImage, blurImage;
    cvtColor(image, grayImage, COLOR_BGR2GRAY);
    GaussianBlur(grayImage, blurImage, Size(5, 5), 0, 0);
    Mat edges;
    Canny(blurImage, edges, 50, 150);
    Mat roiMask = Mat::zeros(image.size(), image.type());
    Point points[4] = {Point(0, image.rows), Point(image.cols / 2, image.rows / 2), Point(image.cols / 2 + 50, image.rows / 2), Point(image.cols, image.rows)};
    fillConvexPoly(roiMask, points, 4, Scalar(255, 0, 0));
    Mat roiImage;
    bitwise_and(edges, roiMask, roiImage);
    vector<Vec2f> lines;
    HoughLines(roiImage, lines, 1, CV_PI / 180, 100, 0, 0);
    vector<Vec2f> leftLines, rightLines;
    for (int i = 0; i < lines.size(); i++) {
        float rho = lines[i][0], theta = lines[i][1];
        if (theta < CV_PI / 4 || theta > 3 * CV_PI / 4) {
            leftLines.push_back(lines[i]);
        } else {
            rightLines.push_back(lines[i]);
        }
    }
    Vec2f leftLine, rightLine;
    if (leftLines.size() > 0) {
        float leftRho = 0, leftTheta = 0;
        for (int i = 0; i < leftLines.size(); i++) {
            leftRho += leftLines[i][0];
            leftTheta += leftLines[i][1];
        }
        leftRho /= leftLines.size();
        leftTheta /= leftLines.size();
        leftLine = Vec2f(leftRho, leftTheta);
    }
    if (rightLines.size() > 0) {
        float rightRho = 0, rightTheta = 0;
        for (int i = 0; i < rightLines.size(); i++) {
            rightRho += rightLines[i][0];
            rightTheta += rightLines[i][1];
        }
        rightRho /= rightLines.size();
        rightTheta /= rightLines.size();
        rightLine = Vec2f(rightRho, rightTheta);
    }
    Point leftStart, leftEnd, rightStart, rightEnd;
    if (leftLine[1] != 0) {
        float leftSlope = -1 / tan(leftLine[1]);
        float leftIntercept = leftLine[0] / sin(leftLine[1]);
        leftStart = Point(0, leftIntercept);
        leftEnd = Point(image.cols, leftSlope * image.cols + leftIntercept);
    } else {
        leftStart = Point(leftLine[0], 0);
        leftEnd = Point(leftLine[0], image.rows);
    }
    if (rightLine[1] != 0) {
        float rightSlope = -1 / tan(rightLine[1]);
        float rightIntercept = rightLine[0] / sin(rightLine[1]);
        rightStart = Point(0, rightIntercept);
        rightEnd = Point(image.cols, rightSlope * image.cols + rightIntercept);
    } else {
        rightStart = Point(rightLine[0], 0);
        rightEnd = Point(rightLine[0], image.rows);
    }
    line(image, leftStart, leftEnd, Scalar(0, 0, 255), 2);
    line(image, rightStart, rightEnd, Scalar(0, 0, 255), 2);
    imshow("Lane Detection", image);
    waitKey(0);
    return 0;
}