Lightweight AI-Based Segmentation of Range-Doppler Maps for Real-Time Selection of Adaptive Processing Algorithms in Airborne Radar Systems

Modern airborne radar systems rely on adaptive signal processing to maintain accurate target detection despite clutter, noise, and active interferences, while operating under tight computational constraints. Conventional strategies for selecting processing algorithms like Moving Target Indicator (MTI), Adaptive Beamforming (ABF), or Space-Time Adaptive Processing (STAP) are often static, limiting effectiveness when disturbances affect only localized regions. This work proposes a lightweight, data-driven segmentation method that analyzes range-Doppler maps to identify distinct spectral regimes and guide regionspecific adaptation of processing. The model combines physically grounded covariance simulations with an efficient convolutional architecture optimized for real-time embedded deployment. By focusing resources where they are most needed, this approach enables more flexible, efficient cognitive radar systems that adjust processing dynamically to complex environments.