Question

当我阅读

的代码时

cv::fisheye::calibrate(InputArrayOfArrays objectPoints, InputArrayOfArrays imagePoints, const Size& image_size,
                                InputOutputArray K, InputOutputArray D, OutputArrayOfArrays rvecs, OutputArrayOfArrays tvecs,
                                int flags , cv::TermCriteria criteria),

我注意到，代码是在旋转矢量和平移矢量的方向上计算出来的，但是高斯-牛顿法的迭代并没有利用它们。

Mat JJ2_inv, ex3;
ComputeJacobians(objectPoints, imagePoints, finalParam, omc, Tc,    check_cond,thresh_cond, JJ2_inv, ex3);
Mat G =  alpha_smooth2 * JJ2_inv * ex3;
currentParam = finalParam + G;

但是currentParam仅包含fx，fy，cx，cy，k1，k2，k3，k4和Alpha

cv::internal::IntrinsicParams cv::internal::IntrinsicParams::operator+(const Mat& a)
{
    IntrinsicParams tmp;
    const double* ptr = a.ptr<double>();

    int j = 0;
    tmp.f[0]    = this->f[0]    + (isEstimate[0] ? ptr[j++] : 0);
    tmp.f[1]    = this->f[1]    + (isEstimate[1] ? ptr[j++] : 0);
    tmp.c[0]    = this->c[0]    + (isEstimate[2] ? ptr[j++] : 0);
    tmp.alpha   = this->alpha   + (isEstimate[4] ? ptr[j++] : 0);
    tmp.c[1]    = this->c[1]    + (isEstimate[3] ? ptr[j++] : 0);
    tmp.k[0]    = this->k[0]    + (isEstimate[5] ? ptr[j++] : 0);
    tmp.k[1]    = this->k[1]    + (isEstimate[6] ? ptr[j++] : 0);
    tmp.k[2]    = this->k[2]    + (isEstimate[7] ? ptr[j++] : 0);
    tmp.k[3]    = this->k[3]    + (isEstimate[8] ? ptr[j++] : 0);

    tmp.isEstimate = isEstimate;
    return tmp;
}

谁能帮助我？

为什么opencv用这种方式计算Extrinsics参数？

0 个答案: