Low-resolution ship classification method of double-flow feature learning generative adversarial network

Abstract

The invention discloses a low-resolution ship classification method of a double-flow feature learning generative adversarial network. The method comprises the steps of 1, establishing a ship image training set; 2, constructing and training a double-flow-channel image decomposer integrating a high-frequency image decomposer and a low-frequency image decomposer; 3, optimizing a generative adversarial network based on feature learning; 4, training a ship classifier; and 5, decomposing, enhancing, splicing and classifying high-frequency and low-frequency components of the low-resolution ship image. Aiming at the problem of lack of low-resolution ship image information, the invention provides a ship image classification method of a double-flow feature learning generative adversarial network, solves the problem of inconsistent loss of a high-frequency component and a low-frequency component in a down-sampling process of a ship image, constructs a double-flow channel image decomposer through high-resolution image guidance. And enhanced image features are generated, almost all input image contents are reserved through image splicing, a low-resolution ship classification task is completed, and the classification effect is good.

Description

technical field [0001] The invention belongs to the technical field of low-resolution ship classification, and in particular relates to a low-resolution ship classification method based on dual-stream feature learning and generation confrontation network. Background technique [0002] The development of object recognition has been greatly promoted by deep learning techniques, such as ResNet, DenseNet and SeNet. Abstract representation and classification of regions of interest. Such models are able to handle images with rich details, but perform poorly with very low-resolution objects. However, distant objects are ubiquitous in many computer vision applications, including satellite Earth-view observations, UAV video surveillance systems, and privacy-preserving video analytics. [0003] Low-resolution ship classification is to classify low-resolution ship images. This task is generally considered to be a very challenging task, because the low-resolution images themselves ha...

AI Technical Summary

Problems solved by technology

The defect of this type of method is how to construct an effective common subspace so that the relative distance of the projected samples is close to the original space

Images