Accurate Optical Target Pose Determination For Applications In Aerial Photogrammetry

We propose a new design for an optical coded target based on concentric circles and a position and orientation determination algorithm optimized for high distances compared to the target size. If two ellipses are fitted on the edge pixels corresponding to the outer and inner circles, quasi-analytical methods are known to obtain the coordinates of the projection of the circles center. We show the limits of these methods for quasi-frontal target orientations and in presence of noise and we propose an iterative refinement algorithm based on a geometric invariant. Next, we introduce a closed form, computationally inexpensive, solution to obtain the target position and orientation given the projected circle center and the parameters of the outer circle projection. The viability of the approach is demonstrated based on aerial pictures taken by an UAV from elevations between 10 to 100 m. We obtain a distance RMS below 0.25 % under 50 m and below 1 % under 100 m with a target size of 90 cm, part of which is a deterministic bias introduced by image exposure.