Parasitic Multistatic Radar (PMR) systems are to equip the world of air traffic surveillance with a reliable and highly cost-effective class of radar systems with counter-stealth abilities. But the computational intensity of the signal processing chain made the process extremely time consuming and acted as the prime hindrance in converting the research project into a practical air surveillance system. Parallel processing using General Purpose Graphic Processing Units (GPGPUs) is used as the solution to handle this computational intensity. The parallel structure of radar signal processing chain with large volume of data fits ideally into the parallel architecture of GPGPUs.
This dissertation details the implementation of PMR signal processing chain in the GPGPU platform. The primary objective of the project is to accelerate the signal processing chain without compromising the algorithm efficiency and to prove that GPGPUs are a promising platform for parasitic radar signal processing. Two distinct clutter cancellation algorithms are implemented together with a high performance matched filter in the GPU platform. The two clutter cancellation algorithm are compared based on their computational and clutter cancellation efficiency and a conclusion about the preferred algorithm for PMR system is made. The GPU implementation of the signal processing chain is compared with the CPU implementation using standard performance metrics focusing on individual stages and the overall system, illustrating the effective acceleration achieved. The dissertation concludes with scope and recommendations for further improvement of the system in a multi-CPU multi-GPU distributed system.