A sar self-focusing method based on image classification

Abstract

The invention discloses a SAR self-focusing method based on image classification, the steps of which include: step S1: acquiring a coarse focus image or a defocus image; step S2: using a VGG network to classify the coarse focus image or defocus image for the first time ; Step S3: Judging the classification result after the VGG classification and performing a second classification; Step S4: Process the image using the autofocus method, and obtain a focused SAR image. The invention has the advantages of simple principle, convenient operation, higher accuracy, better versatility and the like.

Description

technical field [0001] The invention mainly relates to the technical field of airborne synthetic aperture radar, in particular to a SAR self-focusing method based on image classification. Background technique [0002] Airborne synthetic aperture radar (synthetic aperture radar, hereinafter referred to as SAR) can obtain high-resolution two-dimensional images, which can be used in battlefield detection, terrain mapping, disaster prevention and mitigation and other fields. [0003] Airborne SAR achieves high resolution in azimuth by synthesizing the virtual aperture of the antenna through uniform linear motion. However, the aircraft operates in the atmosphere and cannot be guaranteed to fly according to the ideal uniform linear trajectory due to atmospheric disturbances, that is, there are motion errors in the actual trajectory of the aircraft. That is, due to the influence of atmospheric disturbance, the aircraft will deviate from the ideal track, that is, there will be motio...

AI Technical Summary

Problems solved by technology

The parametric method is inefficient, but has low requirements on the image signal-to-noise ratio; the non-parametric method has high efficiency, but has high requirements on the image signal-to-noise ratio, and is generally based on the analysis of strong scattering points

Currently, there is no definite judgment standard, generally relying on the experience of engineers to select methods and set parameters; this also reduces the versatility of general processing methods, requiring program changes according to different scenarios and data

[0010] (2) There is no clear judgment standard for imaging scenes. It is more appropriate to use parametric or non-parametric methods for different scenes. At present, there is no judgment standard, but only a vague empirical judgment

[0011] (3) The imaging algorithm is not universal. Due to the lack of judgment standards, the imaging algorithm needs to be manually and repeatedly debugged, and different methods and parameters need to be selected for different data. The algorithm is not universal

Images