Rapid frequency-domain backward projection three-dimensional imaging method based on GPU

Abstract

The invention discloses a rapid frequency-domain backward projection three-dimensional imaging method based on a GPU. The method is characterized by based on a parallel processing ability of the GPU,using a three-dimensional frequency-domain backward projection algorithm, using a graphic processing unit GPU and adopting a single-program multi-data instruction mode to schedule abundant hardware resources so that one thread in the graphic processing unit GPU is responsible for one frequency spectrum grid point; in each thread, calculating backward projection values of current frequency domain grid points in all array directions in parallel and accumulating; and finally, through three-dimensional inverse Fourier transform, acquiring a three-dimensional SAR image. Operation efficiency of a backward projection algorithm is greatly increased. Compared with the prior art, the method possesses high operation efficiency and a three-dimensional imaging characteristic of a linear array SAR is realized. The method is suitable for synthetic aperture radar imaging, earth remote sensing and other fields.

Description

Technical field: [0001] The technical invention belongs to the technical field of radar, and in particular relates to the technical field of synthetic aperture radar (SAR) imaging. Background technique: [0002] Due to its advantages of all-day, all-weather and large-scale scene observation, synthetic aperture radar (SAR) has become an important remote sensing technology for large-area topographic surveying and mapping. greater effect. Linear array SAR (LASAR) is an extension of the target dimension space resolution capability of traditional two-dimensional SAR imaging technology, which can obtain three-dimensional radar imaging of the observed target scene, and can more precisely describe the geometric and scattering characteristics of the target in the observed scene, compared with the traditional two-dimensional SAR imaging technology. Two-dimensional SAR has improved the target feature extraction and target recognition capabilities of radar systems, and has become a hot...

AI Technical Summary

Problems solved by technology

Although the frequency domain BP algorithm has a greater efficiency improvement compared with the time domain BP algorithm, it still takes more time in the case of a large scene, because as the scene increases, the number of grid points that need to be divided during imaging It increases exponentially, and the backprojection value needs to be calculated one by one in the imaging process, and the calculation amount is extremely huge, which limits its practical application

Images