A sar image classification method based on feature-level statistical description learning

Abstract

The invention discloses a SAR image classification method for feature-level statistical description learning, which includes: inputting a target SAR image into a SAR image classification network; The feature primitives of the middle layer semantics extract the statistical primitive vector of the target SAR image; the nonlinear and linear transformation layers generate the feature-level statistical description vector of the target SAR image based on the statistical primitive vector; the Softmax layer generates the target SAR image based on the feature-level statistical description vector classification results. Compared with the traditional CNN method, the present invention not only focuses on the structural feature learning of SAR images, but also considers the feature-level statistical characteristics of SAR images in the feature learning process, and is committed to integrating feature learning and statistical analysis into one, which can effectively solve the problem of Insufficient generalization ability of CNN method for SAR image classification.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a SAR image classification method for feature-level statistical description learning. Background technique [0002] Synthetic Aperture Radar (SAR) can obtain rich information of ground objects all day and all day long, but its coherent imaging mechanism makes SAR images present inherent random image patterns, so the statistical analysis of SAR images is very important for geography and biology. The extraction of such information is very important. Under the coherent imaging mechanism, each pixel of the SAR image is formed by the coherent superposition of echoes from multiple scattering centers in the corresponding resolution unit. Due to the randomness of the location distribution of the scattering centers in the resolution unit, the echoes of the scattering centers have random phases, so the echoes of multiple scattering centers are also random after coherent superposi...

AI Technical Summary

Problems solved by technology

[0005] For above-mentioned deficiencies in the prior art, the problem that the present invention needs to solve actually is: the generalization ability is insufficient when utilizing CNN method to carry out SAR image classification

Images