A Robust Cognitive Electronic Support Measures Approach Based on Computer Vision

In electronic warfare, a robust perception of the surrounding signal environment is vital for optimal countermeasures. In this article, a new approach to a cognitive electronic support measures (COG-ESM) system based on computer vision techniques is proposed. The proposed approach overcomes the limitations of the traditional ESM when subjected to complex radar signals in a dense environment. The proposed approach relaxes the deinterleaving process of the concurrent radar signals to a 2D emitter localization in the time-frequency (TF) domain. Also, it integrates the modulation classification process with the 2D emitter localization in one process. This merge allows for sharing the hardware and software assets required for both processes. The proposed COG-ESM applies the state-of-the-art object detection networks to the TF images to localize and classify the individual radar emitters. The COG-ESM is trained to perform both 2D localization and modulation classification under the multipath effect, hence it is more robust under channel impairments. A new dataset of complex interleaved radar signals representing a dense signal environment under the multipath channel effect is introduced. Typical scenarios, signal parameters, and modulation types are considered in generating the dataset. That dataset is called wideband radar signals in dense environment (WBR-DE).