我正在使用Levenberg-Marquardt algorithm最小化一个包含6个参数的非线性函数。对于每个最小化,我有大约50个数据点,但我没有得到足够准确的结果。我的参数彼此相差几个数量级,这一事实真的如此重要吗?如果是,我应该在哪里寻找解决方案?如果没有,你在工作中遇到了LMA的哪些限制(它可能有助于发现我的应用程序的其他问题)?非常感谢你的帮助。
编辑:我试图解决的问题是确定最佳转换T:
typedef struct
{
double x_translation, y_translation, z_translation;
double x_rotation, y_rotation, z_rotation;
} transform_3D;以将3D点集拟合到3D线束。详细地说,我得到了一组三维点的坐标和相应的三维线的方程,它们应该通过这些点(在理想情况下)。LMA正在最小化变换的3D点到相应3D线的距离总和。转换函数如下:
cv::Point3d Geometry::transformation_3D(cv::Point3d point, transform_3D transformation)
{
cv::Point3d p_odd,p_even;
//rotation x
p_odd.x=point.x;
p_odd.y=point.y*cos(transformation.x_rotation)-point.z*sin(transformation.x_rotation);
p_odd.z=point.y*sin(transformation.x_rotation)+point.z*cos(transformation.x_rotation);
//rotation y
p_even.x=p_odd.z*sin(transformation.y_rotation)+p_odd.x*cos(transformation.y_rotation);
p_even.y=p_odd.y;
p_even.z=p_odd.z*cos(transformation.y_rotation)-p_odd.x*sin(transformation.y_rotation);
//rotation z
p_odd.x=p_even.x*cos(transformation.z_rotation)-p_even.y*sin(transformation.z_rotation);
p_odd.y=p_even.x*sin(transformation.z_rotation)+p_even.y*cos(transformation.z_rotation);
p_odd.z=p_even.z;
//translation
p_even.x=p_odd.x+transformation.x_translation;
p_even.y=p_odd.y+transformation.y_translation;
p_even.z=p_odd.z+transformation.z_translation;
return p_even;
}希望这个解释能有所帮助……
Edit2:
下面粘贴了一些示例数据。三维线由中心点和方向向量描述。所有直线的中心点是(0,0,0),每个矢量的'uz‘坐标等于1。方向矢量的'ux’坐标集:
-1.0986, -1.0986, -1.0986,
-1.0986, -1.0990, -1.0986,
-1.0986, -1.0986, -0.9995,
-0.9996, -0.9996, -0.9995,
-0.9995, -0.9995, -0.9996,
-0.9003, -0.9003, -0.9004,
-0.9003, -0.9003, -0.9003,
-0.9003, -0.9003, -0.8011,
-0.7020, -0.7019, -0.6028,
-0.5035, -0.5037, -0.4045,
-0.3052, -0.3053, -0.2062,
-0.1069, -0.1069, -0.1075,
-0.1070, -0.1070, -0.1069,
-0.1069, -0.1070, -0.0079,
-0.0079, -0.0079, -0.0078,
-0.0078, -0.0079, -0.0079,
0.0914, 0.0914, 0.0913,
0.0913, 0.0914, 0.0915,
0.0914, 0.0914方向向量的'uy‘坐标集:
-0.2032, -0.0047, 0.1936,
0.3919, 0.5901, 0.7885,
0.9869, 1.1852, -0.1040,
0.0944, 0.2927, 0.4911,
0.6894, 0.8877, 1.0860,
-0.2032, -0.0047, 0.1936,
0.3919, 0.5902, 0.7885,
0.9869, 1.1852, 1.0860,
0.9869, 1.1852, 1.0861,
0.9865, 1.1853, 1.0860,
0.9870, 1.1852, 1.0861,
-0.2032, -0.0047, 0.1937,
0.3919, 0.5902, 0.7885,
0.9869, 1.1852, -0.1039,
0.0944, 0.2927, 0.4911,
0.6894, 0.8877, 1.0860,
-0.2032, -0.0047, 0.1935,
0.3919, 0.5902, 0.7885,
0.9869, 1.1852和(x.y.z.)中的三维点的集合表格:
{{0, 0, 0}, {0, 16, 0}, {0, 32, 0},
{0, 48, 0}, {0, 64, 0}, {0, 80, 0},
{0, 96, 0}, {0, 112,0}, {8, 8, 0},
{8, 24, 0}, {8, 40, 0}, {8, 56, 0},
{8, 72, 0}, {8, 88, 0}, {8, 104, 0},
{16, 0, 0}, {16, 16,0}, {16, 32, 0},
{16, 48, 0}, {16, 64, 0}, {16, 80, 0},
{16, 96, 0}, {16, 112, 0}, {24, 104, 0},
{32, 96, 0}, {32, 112, 0}, {40, 104, 0},
{48, 96, 0}, {48, 112, 0}, {56, 104, 0},
{64, 96, 0}, {64, 112, 0}, {72, 104, 0},
{80, 0, 0}, {80, 16, 0}, {80, 32, 0},
{80,48, 0}, {80, 64, 0}, {80, 80, 0},
{80, 96, 0}, {80, 112, 0}, {88, 8, 0},
{88, 24, 0}, {88, 40, 0}, {88, 56, 0},
{88, 72, 0}, {88, 88, 0}, {88, 104, 0},
{96, 0, 0}, {96, 16, 0}, {96, 32, 0},
{96, 48,0}, {96, 64, 0}, {96, 80, 0},
{96, 96, 0}, {96, 112, 0}} 这是一种具有非常小旋转的“简单”建模数据。
发布于 2010-12-14 16:49:24
使用Levenberg-Marquardt的正确方法是,您需要一个良好的参数初始估计(“种子”)。回想一下,LM是Newton-Raphson的变体;与此类迭代算法一样,起点的质量将决定迭代的成败;要么收敛到您想要的东西,要么收敛到完全不同的东西(并不是那么不可能发生,特别是如果您有很多参数的话),或者飞向远处的狂野(分歧)。
在任何情况下,如果您可以提到您正在拟合的模型函数,以及可能的数据散点图,将会更有帮助;这可能会对找到一个可行的解决方案大有帮助。
发布于 2010-12-14 21:31:56
我建议您尝试使用一种不同的方法来间接查找旋转参数,即使用4x4 affine transformation matrix合并平移和旋转参数。
这消除了正弦和余弦函数的非线性(您可以在事后弄清楚)。
困难的部分是限制变换矩阵的剪切或缩放,这是你不想要的。
发布于 2015-12-26 01:46:04
好吧,我使用ceres-solver解决了这个问题,但我确实对您的数据进行了修改。我没有使用"uz=1.0",而是使用了"uz=0.0“,这使得这完全是一个2d数据拟合。
我得到了以下结果。传输:-88.6384,-16.3879,0旋转: 0,0,-6.97813e-05
在得到这些结果后,手动计算变换后的点到相应直线的正交距离之和,得到0.0280452。
struct CostFunctor {
CostFunctor(const double p[3], double ux, double uy){
p_[0] = p[0];p_[1] = p[1];p_[2] = p[2];
n_[0] = ux; n_[1] = uy;
n_[2] = 0.0;
normalize(n_);
}
template <typename T>
bool operator()(const T* const x, T* residual) const {
T pDash[3];
T pIn[3];
T temp[3];
pIn[0] = T(p_[0]);
pIn[1] = T(p_[1]);
pIn[2] = T(p_[2]);
//transform the input point p_ to pDash
xform(x, &pIn[0], &pDash[0]);
//find dot(pDash, n), where n is the direction of line
T pDashDotN = T(pDash[0]) * T(n_[0]) + T(pDash[1]) * T(n_[1]) + T(pDash[2]) * T(n_[2]);
//projection of pDash along line
temp[0] = pDashDotN * n_[0];temp[1] = pDashDotN * n_[1];temp[2] = pDashDotN * n_[2];
//orthogonal vector from projection to point
temp[0] = pDash[0] - temp[0];temp[1] = pDash[1] - temp[1];temp[2] = pDash[2] - temp[2];
//squared error
residual[0] = temp[0] * temp[0] + temp[1] * temp[1] + temp[2] * temp[2];
return true;
}
//untransformed point
double p_[3];
double ux_;
double uy_;
//direction of line
double n_[3];
};
template<typename T>
void xform(const T *x, const T * inPoint, T *outPoint3) {
T xTheta = x[3];
T pOdd[3], pEven[3];
pOdd[0] = inPoint[0];
pOdd[1] = inPoint[1] * cos(xTheta) + inPoint[2] * sin(xTheta);
pOdd[2] = -inPoint[1] * sin(xTheta) + inPoint[2] * cos(xTheta);
T yTheta = x[4];
pEven[0] = pOdd[0] * cos(yTheta) + pOdd[2] * sin(yTheta);
pEven[1] = pOdd[1];
pEven[2] = -pOdd[0] * sin(yTheta) + pOdd[2] * cos(yTheta);
T zTheta = x[5];
pOdd[0] = pEven[0] * cos(zTheta) - pEven[1] * sin(zTheta);
pOdd[1] = pEven[0] * sin(zTheta) + pEven[1] * cos(zTheta);
pOdd[2] = pEven[2];
T xTrans = x[0], yTrans = x[1], zTrans = x[2];
pOdd[0] += xTrans;
pOdd[1] += yTrans;
pOdd[2] += zTrans;
outPoint3[0] = pOdd[0];
outPoint3[1] = pOdd[1];
outPoint3[2] = pOdd[2];
}https://stackoverflow.com/questions/4437318
复制相似问题