TRABALHOS PUBLICADOS
2021 |
Experimental Assessment of the Viability of a GAN Model for Radar Data Generation Proceedings Article Queiroz, Samuel Souza Alcântara; Shiguemori, Élcio Hideiti; Augusto, Carlos Vinícius Souza; Souza, Robson Porto; d’Amore, Roberto; Faria, Lester Abreu Resumo | Links | BibTeX | Tags: FMCW Radar, Generative Adversarial Network @inproceedings{GAN2021Generativeb,In the Defense industry, radar target simulations are a major point of technologic independence for countries, once allows creating and evaluating Radar Signatures of complex targets, as aircraft, ships and armored cars, without the need of measuring them. It is a possibility of generating enemy’s signatures and defining the best approach to react and detect them. However, this kind of simulation is expensive, spend a lot of computational resources and demands a complex setup. This study proposes a proof of concept of a radar target simulation, based on a case study of automotive radars using Generative Adversarial Networks (GAN): an artificial intelligence technique that is used for generating realistic synthetic data. The achieved results show that it can generate realistic radar targets using low computational efforts, opening a new way to radar target simulations: AI-based simulations. Although being based on automotive targets, all results can be extrapolated to Defense scenarios. |