DLT (Direct Linear Transform) algorithm

first, the definition of direct linear transformation method is to establish the "coordinate instrument coordinate" of the image point and the direct linear relation of the object square coordinate of the corresponding point. Features of the direct linear transformation solution:
Non-centring, indefinite items do not require an internal and external azimuth element The starting value of the object square space need to arrange a set of control points is particularly suitable for dealing with non-measurement camera photography is a kind of space after the intersection-owe solution.

Second, deduction

(X¯,y¯) [On2,on1] \bar{x}, \bar{y}) [On_2, on_1]\-as the origin of the primary point, does not include the coordinates of the image point P of the linear error;
[Om2,om1] [Om_2, om_1]\-contains the coordinates of the non-orthogonal dβd\beta\ as the origin of the primary point;
[Om2,om1˙] [Om_2, o\dot{m_1}]\-the coordinates of the image point P with the non-orthogonal dβd\beta\ error and the scale error DS as the origin point of the primary point;

The above hypothesis assumes that the x-axis direction is not affected by scale error. With the x-axis scaling factor of 1, the y-axis Direction scale factor is (1 + DS), and the x-axis direction is the main pitch is FX f_x \, the y-axis direction is
fy=fx/(1+ds) f_y = f_x/(1 + DS) \;

Δx=on2−om2=m2p⋅sindβ=om1⋅sindβ= (1+ds) (y−y0) ⋅sindβ≈ (y−y0) ⋅sindβ\delta x = on_2-om_2 = M_2p\cdot Sind\beta = om_1 \cdot s Ind\beta = (1+ds) (y-y_0) \cdot Sind\beta \approx (y-y_0) \cdot Sind\beta