UAV-based marine mammals positioning and tracking system

In this paper, a new strategy to localize and track marine mammals moving at the ocean surface is proposed, resorting to measurements provided by a GPS, an Attitude and Heading Reference System (AHRS), and images provided by a camera installed onboard an Unmanned Aerial Vehicle (UAV). The segmentation of the marine mammals in the images is tackled resorting to methodologies based on active contours. The measurements of the position of the target are combined with data from the position and attitude of the UAV, provided by the GPS and AHRS, respectively, leading to target position estimates, in an inertial reference frame. To obtain these estimates, two Kalman Filters are proposed: i) a time-invariant filter, that estimates only the target position, and ii) a time-varying filter, that combines estimates of the position of the target with those of the position of the UAV, which is shown to improve the overall performance. Simulation results illustrating the behavior of the system in realistic conditions are presented and discussed. Results are also reported in the case where occlusions in the images occur, namely during the periods where the marine mammals dive, to allow the filter robustness assessment.