Robust aerial scene-matching algorithm based on relative velocity model

Abstract

We present a robust scene-matching (SM) algorithm using time-invariant features that are propagated and bounded by a model propagator and pixel boundary. The SM based absolute navigation has the advantage that the position of the vehicle can be independently calculated without external information, making it possible to calculate a stable navigation solution without cumulative errors. However, SM-based absolute localization has a mismatching problem, this is due to the difference between the reference for the matching and the input image, and the more the change, the higher the probability of mismatching. In this paper we propose an algorithm that can mitigate the mismatching problem with a model-based propagator and time-invariant features. The propagator is based on a relative velocity of the inertial navigation system (INS) model, which is very accurate for a short time. Also the propagated feature points have pixel boundaries, which considers not only INS model uncertainty but also distortion of the aerial images caused by various terrain characteristics. The proposed algorithm is verified by simulation using real experimental data. Consequently we can found the proposed algorithm is very effective in mitigating the mismatching problem in urban areas.