Robust tracking of vessels in oceanographic airborne images

In this paper we present and evaluate an algorithm for tracking vessels in oceanographic airborne image sequences on the visible spectrum. Such sequences are challenging due to sun reflections, wakes, wave crests and fast motions, which significantly degrade the performance of general purpose tracking algorithms. The proposed method is based on state-of-the-art correlation filter tracking complemented with an image segmentation and blob analysis stage. The purpose of this later stage is to re-center the target in the tracking window to compensate for drifts in the correlation filter. We evaluate our proposal using a known benchmark in the field and compare it with general purpose tracking algorithms. Results show that our method beats the general purpose state-of-the-art tracking algorithms in the airborne maritime scenario both in performance and in computation time.