原理参考：直线拟合——最小二乘法、hough变换
/ * @description: 直线拟合 * @param points 输入点集 * @param line_para 直线参数 */ void fitline(std::vector<cv::Point>& points, cv::Vec4f& line_para) {
int N = points.size(); if (N < 2) return; double sum_x = 0, sum_y = 0, sum_xx = 0, sum_xy = 0, sum_yy = 0; for (int i = 0; i < N; i++) {
sum_x += points[i].x; sum_y += points[i].y; sum_xx += points[i].x * points[i].x; sum_xy += points[i].x * points[i].y; sum_yy += points[i].y * points[i].y; } //least squares: y=kx+b float k = (N*sum_xy-sum_x*sum_y) / (N*sum_xx - sum_x*sum_x); float b = (sum_xx*sum_y - sum_x*sum_xy) / (N*sum_xx - sum_x*sum_x); std::cout << "k = " << k << ", b = " << b << std::endl; //total least squares: ax+by+c=0 float A, B, C; float mean_x = sum_x / N, mean_y = sum_y / N, mean_xx = sum_xx / N, mean_xy = sum_xy / N, mean_yy = sum_yy / N; cv::Mat m = (cv::Mat_<float>(2, 2) << mean_xx - mean_x*mean_x, mean_xy - mean_x*mean_y, mean_xy - mean_x*mean_y, mean_yy - mean_y*mean_y); cv::Mat eigenvalue, eigenvector; cv::eigen(m * N, eigenvalue, eigenvector); float v0 = eigenvalue.at<float>(0, 0), v1 = eigenvalue.at<float>(0, 1); if (abs(v0) < abs(v1)) {
A = eigenvector.at<float>(0, 0); B = eigenvector.at<float>(0, 1); } else {
A = eigenvector.at<float>(1, 0); B = eigenvector.at<float>(1, 1); } C = -(A*mean_x + B*mean_y); std::cout << "k = " << -A / B << ", b = " << -C / B << std::endl; line_para[0] = B; line_para[1] = -A; line_para[2] = mean_x; line_para[3] = mean_y; } 代码传送门：https://github.com/taifyang/OpenCV-algorithm
版权声明：本文内容由互联网用户自发贡献，该文观点仅代表作者本人。本站仅提供信息存储空间服务，不拥有所有权，不承担相关法律责任。如发现本站有涉嫌侵权/违法违规的内容， 请联系我们举报，一经查实，本站将立刻删除。
发布者：全栈程序员-站长，转载请注明出处：https://javaforall.net/211725.html原文链接：https://javaforall.net
