A SAR Image Sidelobe Removal Method Based on Generative Adversarial Neural Network

Abstract

The invention discloses a SAR sidelobe removal method based on a generative adversarial neural network. A sidelobe removal network is constructed based on the pix2pix model of the generative adversarial network. The point target map is used as the label of the network, and the result map without sidelobe generated based on the generative confrontation network is used as the output of the network; the point target map without sidelobe is obtained by assigning different intensity values ​​to the random point target, and the sidelobe-free The point target image convolution two-dimensional shock response function generates the corresponding point target image with side lobes, and constructs the simulation data training set; uses the simulation data to train the side lobe network model; uses the trained side lobe network model to SAR images are sidelobe removed. The present invention has a certain inhibitory effect on noise and defocus while improving the side lobe removal effect.

Description

technical field [0001] The invention relates to a synthetic aperture radar image sidelobe suppression processing technology, in particular to a SAR (Synthetic Aperture Radar) image sidelobe removal method based on a generative adversarial neural network. Background technique [0002] Synthetic Aperture Radar (SAR) is a high-resolution imaging radar, which plays an important role in the fields of ocean, disaster reduction and national defense. During SAR imaging, it is necessary to perform matched filtering on the echo data in the range and azimuth directions, so that the shock response of an ideal point target is a two-dimensional sinc function, so the target point will have side lobes after imaging. Since the side lobe of a strong scattering target may be higher than the host level of a weak target, the weak target will be covered, which is not conducive to SAR image interpretation and target detection and recognition. In particular, for targets such as ships and vehicles,...

AI Technical Summary

Problems solved by technology

[0005] 1) The frequency domain weighting method is a windowing operation on the signal spectrum. While suppressing the side lobe, it broadens the main lobe of the signal, which will lead to a decrease in resolution

[0006] 2) The amplitude of the main lobe sampling point of the image processed by other weighted windows in the double orbital side lobe suppression method is smaller than the first side lobe amplitude of the image processed by the rectangular window. Although the side lobe is suppressed to a certain extent, it is also It will cause the first peak side lobe to appear in advance, and the ability to suppress the distant side lobe is limited

[0007] 3) The spatial apodization sidelobe suppression method can suppress the sidelobe level without losing the image resolution, but for the stray target of the SAR image, this method may lead to the loss of sponsor energy

[0008] 4) The existing three main processing methods are to suppress the side lobe as much as possible, but cannot completely eliminate the side lobe

Images