Parallel implementation of real-time ISAR imaging based on graphics processing units

Abstract

Inverse Synthetic aperture radar imaging is widely utilized in many fields. But high resolution ISAR imaging of maneuvering targets is challenging due to its complexity during long coherence processing interval (CPI). So the ISAR imaging of maneuvering targets needs to be processed in short CPI. The resolution in cross-range is inversely proportional to the synthesized aperture length. So the RD imaging is blurred in short CPI. Compressive sensing (CS) theory indicates that it is possible to obtain precise recovery of a sparse signal from very limited measurements, which can improve imaging resolution of short aperture. But the CS results in a high computational complexity. Therefore, the high computational complexity of CS is a major concern for its implementation to achieve real-time reconstruction of compressively sensed signals. One of recovery algorithms for CS is the orthogonal matching pursuit (OMP), which involves massive matrix/vector operations. The OMP algorithm is very suits to be implemented in parallel on Graphics processing units (GPU) platform to accelerate the computation. The traditional GPU board has large size and high power consumption, which limits its application in the field of radar signal processing. In this paper, a real-time ISAR imaging conceptive is presented based on embedded GPU platform. Experiment results show that we can apply the CS imaging to achieve higher resolution than RD imaging in the case of short CPI. What’s more, remarkable speedup is achieved by our implementation of OMP on embedded GPU than CPU.