Atomic Decomposition-Based SAR Imaging And Its Implementation Technologys Based On GPU

The classic SAR imaging algorithms , which focus the SAR raw data as a whole, are developed based on specified imaging geometry model between the detected scene and the flying platform. With them, the content of the final image cannot be controlled. Furthermore, due to the difference between the pre-assumed geometry models and the actual one, the focusing parameters and the actual data cannot exactly match, to a certain extent, leading to defocusing and position deviation of image data. With the development of SAR technique, the spatial resolution becomes even finer, and the latest resolution even reached 0.1m. Such high resolution requires a more accurate geometry model. Considering higher resolution and more accurate geometric model, the traditional algorithms become more and more complicated.To solve the above-mentioned problems, in this paper, we propose a atomic decomposition-based SAR imaging algorithm, which is independe- nt of geometric model. This algorithm utilizes AD to estimate focusing parameters from raw data or the range compressed data. With these parameters, the targets or scatter-points in the scene will be accurately imaged progressively. With the new processing scheme, the targets or scat- tering points can be focused equally in spite of their motion state, and the content of the final image can be controlled. However, the efficiency of the AD-based SAR imaging technique is low due to AD process. In this paper, we transfer the AD process, which is the key part of the new algorithm, to GPU to enhance the computation performance.The main innovations of this paper are as follows:1. To propose AD-based SAR imaging technique. There are detailed theoretical description and process analysis of the new algorithm. Then we compare the AD-based SAR imaging technique with the classic SAR imaging algorithm and list the advantages of the new algorithm. Of course, the efficiency of AD-based SAR imaging technique is low. Finally, some methods or skill have been adopted to enhance the computation performance.2. To propose a highly efficient GPU-based method for decomposing SAR azimuth signal. In this paper, we describe the process of this method. Besides, algorithm optimization according to the CUDA model is presented. Finally, we compare the GPU-based method for decomposing SAR azimuth signal with the traditional CPU-based one.