Extended Drones Tracking From ISAR Images With Doppler Effect and Orientation Based Robust Sub-Random Matrices Algorithm

The simultaneous imaging and tracking of drones and small UAVs is one of the most challenging problems in inverse synthetic aperture radar (ISAR) signal processing and has received significant attention recently. Wide-angle ISAR imaging may offer better resolution and richer geometrical features. Aiming at this issue, we propose a new algorithm for multiple extended drone targets (EDTs) detection and tracking from wide-angle ISAR images. Since a conventional tracking algorithm using a random matrices model (RMM) is not accurate enough to reconstruct a drone's extension by one ellipse, an appropriate technique needs to be developed for ISAR imaging and tracking of EDTs. This paper models the EDT's extension as multiple ellipses with a time-varying orientation angle and solves the resulting interference problem involving data association between the nonlinear measurements and sub-ellipses. Recently, the multi-Bernoulli (MB) filter for extended targets has been proposed with the RMM approach for linear ellipsoidal target tracking by additional state variables. However, the implementation of the MB-RMM filter for nonlinear non-Gaussian models and the Doppler effect has not been presented, especially for ISAR images. We consider linearizing the nonlinear observations to solve the nonlinearity due to the Doppler effect, target motion, and wide-angle ISAR images. We presented a new robust Sub-RMM-MB-TBD filter to obtain the estimated orientation of EDT's shape after using the range-only motion compensation that uses both range and Doppler information. Moreover, the authors applied the k-space algorithm to generate ISAR image. Simulation results demonstrate this remarkable performance.