Multitask Learning for Radar-Based Characterization of Drones
A. Pappas (TU Delft - Microwave Sensing, Signals & Systems)
J.J.M. de Wit (TNO, TU Delft - Microwave Sensing, Signals & Systems)
F. Fioranelli (TU Delft - Microwave Sensing, Signals & Systems)
Bas Jacobs (TNO)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
For the effective deployment of countermeasures against drones, information on their intent is crucial. There are several indicators for a drone's intent, e.g., its size, payload, and behavior. Within the current study, the focus was on estimating subsets of the following four indicators: a drone's wing type, its number of rotors, the presence of a payload and its mean rotor rotation rate. Three Multitask Learning (MTL) approaches were analyzed for the simultaneous estimation of subsets of these indicators based on radar micro-Doppler spectrograms. MTL refers to training neural networks simultaneously for multiple related tasks. The assumption is that if tasks share features between them, an MTL model is easier to train and has improved generalization capabilities as compared to separately trained single-task neural networks. The results of this initial study show that MTL provides overall better performance than the single-task learning approach, given the available data set of measured drone spectrograms.