C++：最小二乘法拟合直线_pcl最小二乘拟合空间直线c++-CSDN博客

本文链接：https://blog.csdn.net/xinjiang666/article/details/103782544

本文分享了两种使用最小二乘法进行直线拟合的方法，包括斜截式y=kx+b和一般式Ax+By+C=0。通过具体算法实现，详细介绍了如何从数据点集中找出最佳拟合直线。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

分享给有需要的人，代码质量勿喷。

一、y = k * x + b

bool xjLeastSquares::xjFitLineByLeastSquares(
std::vector<double> &xjParameters, const std::vector<xjPoint> &xjData,
	const bool &kxb)
{
	try
	{
		xjParameters.clear();
		if (xjData.size() < 2)
			return false;

		double k = 0, b = 0;

		int xjPcount = xjData.size();
		double sumX = 0.0, sumY = 0.0;
		double sumXX = 0.0, sumXY = 0.0;
		for (int i = 0; i < xjPcount; i++)
		{
			sumX += xjData[i].x;
			sumY += xjData[i].y;

			sumXX += xjData[i].x * xjData[i].x;
			sumXY += xjData[i].x * xjData[i].y;
		}
		double deltaX = sumXX * xjPcount - sumX * sumX;

		if (std::abs(deltaX) > 1e-15)/* 防止竖直线 */
		{
			k = (sumXY*xjPcount - sumX * sumY) / deltaX;
			b = (sumXX*sumY - sumX * sumXY) / deltaX;
		}

		xjParameters.push_back(k);
		xjParameters.push_back(b);

		return true;
	}
	catch (const std::exception&)
	{
		return false;
	}
}

二、AX + BY + C = 0

参考自 Ziv Yaniv

/* 最小二乘法拟合直线：AX+BY+C=0 */
bool xjLeastSquares::xjFitLineByLeastSquares(
std::vector<double> &xjParameters, const std::vector<xjPoint> &xjData)
{
	try
	{
		xjParameters.clear();
		if (xjData.size() < 2)
			return false;

		double A = 0.0, B = 0.0, C = 0.0;

		int xjPcount = xjData.size();
		double meanX = 0.0, meanY = 0.0;
		double sumXX = 0.0, sumXY = 0.0, sumYY = 0.0;
		for (int i = 0; i < xjPcount; i++)
		{
			meanX += xjData[i].x;
			meanY += xjData[i].y;

			sumXX += xjData[i].x * xjData[i].x;
			sumXY += xjData[i].x * xjData[i].y;
			sumYY += xjData[i].y * xjData[i].y;
		}
		meanX /= xjPcount;
		meanY /= xjPcount;

		sumXX -= xjPcount * meanX*meanX;
		sumXY -= xjPcount * meanX*meanY;
		sumYY -= xjPcount * meanY*meanY;
		if (abs(sumXX) < 1e-15)
		{
			A = 1.0;
			B = 0.0;
		}
		else
		{
			double ev = (sumXX + sumYY + sqrt((sumXX - sumYY)*(sumXX - sumYY) + 4 * sumXY*sumXY)) / 2.0;
			A = -sumXY;
			B = ev - sumYY;
			double norm = sqrt(A*A + B * B);
			A /= norm;
			B /= norm;
		}

		xjParameters.push_back(A);
		xjParameters.push_back(B);
		C = -(A * meanX + B * meanY);
		xjParameters.push_back(C);

		return true;
	}
	catch (const std::exception&)
	{
		return false;
	}
}