SAR three-dimensional rotating target refocusing method based on CV-ConvGRU

Abstract

The invention discloses a CV-ConvGRU-based SAR (Synthetic Aperture Radar) three-dimensional rotating target refocusing method, relates to the technical field of image processing, and provides a CV-ConvGRU for solving the problem of SAR imaging blurring caused by the influence of factors such as sea waves, wind and self navigation in the prior art, the CV-ConvGRU expands a whole network comprising a convolutional layer, an activation function, an update gate and a reset gate to a complex domain, and the CV-ConvGRU is used for realizing the refocusing of a three-dimensional rotating target. And on the basis, a back propagation algorithm used for CV-ConvGRU training is deduced. According to the invention, a CV-SSRN architecture is designed based on the CV-ConvGRU, SAR three-dimensional rotating target refocusing is carried out, an SAR refocusing task is converted into an image regression problem, and an encoder and a decoder are added to realize target refocusing, so that the focusing precision is obviously improved, and the problem of SAR imaging blurring is avoided. According to the method, the SAR imaging principle and the complex field deep learning technology are utilized, the complex SAR refocusing requirement is better met, and the stability of visual angle change, affine transformation and noise is kept to a certain degree.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a CV-ConvGRU-based SAR three-dimensional rotating target refocusing method. Background technique [0002] Synthetic aperture radar (SAR) has indisputable advantages in the fields of military reconnaissance, geographic mapping, and resource measurement because it provides high-resolution images of observation scenes with all-weather and all-day surveillance capabilities. Due to the influence of sea waves, wind, own navigation and other factors, the ship target at sea is generally in a complex three-dimensional swing state. The traditional SAR imaging algorithm is designed for static scene imaging, so in the SAR image obtained by the method based on matched filtering, the complex motion of the target will cause the image blur. Good imaging quality is the premise of subsequent classification and recognition processing. Therefore, when performing SAR image processing, refoc...

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to propose a SAR three-dimensional rotating target refocusing method based on CV-ConvGRU in view of the problem of blurred SAR imaging caused by factors such as sea waves, wind, and self-navigation in the prior art

Images