Performance enhancement of the FRACTA algorithm via dimensionality reduction: results from KASSPER I

The recently developed FRACTA algorithm is a reiterative censoring and detection algorithm for space-time adaptive processing (STAP) which has been shown to provide excellent detection performance in non-homogeneous interference environments. In this paper, a partially adaptive implementation of the FRACTA algorithm is presented which significantly reduces the high computational complexity and large sample-support requirements of fully adaptive STAP. Prior to application of the FRACTA algorithm, the data is passed through a multi-window post-Doppler dimensionality reduction transformation. Two implementations of the multi-window post-Doppler dimensionality reduction methodology are considered: PRI-staggered and adjacent-bin. A detailed analysis of the application of the partially adaptive FRACTA algorithm to the KASSPER I challenge datacube is presented. The PRI-staggered approach with D=6 filters per Doppler bin is found to provide the best detection performance, outperforming the fully adaptive case while simultaneously reducing the runtime by a factor of ten. Using this implementation, partially adaptive FRACTA detects 197 out of 268 targets with one false alarm (the clairvoyant processor detects 198 targets). In addition, the partially adaptive FRACTA algorithm is shown to be resilient to jamming, and performs well for reduced sample support situations. When compared to partially adaptive STAP using traditional sliding window processing (SWP), the runtime of partially adaptive FRACTA is 14 times faster, and the detection performance is significantly increased (SWP detects 46 out of 268 targets with one false alarm).