Small sample remote sensing image scene classification method based on iterative feature distribution learning

Abstract

The invention discloses a small sample remote sensing image scene classification method based on iterative feature distribution learning. The method comprises the following steps: step 1, setting an iterative feature distribution learning network framework; step 2, generating a similarity matrix between the image samples; step 3, training a small sample classifier by utilizing guide knowledge based on classification prediction probability and a similarity matrix; and step 4, correcting the feature distribution by using the prediction probability distribution matrix and the attention weight, and then iteratively updating the whole network. The method is simple in structure and reasonable in design, firstly, the similarity matrix is generated to obtain feature association between samples, secondly, the classifier can be trained under the guidance of historical classification prediction, finally, the attention mechanism is adopted to correct the features, and the corrected features are input into the network again to realize iterative updating of the whole network. According to the method, through iterative feature distribution learning, the characterization capability of the category can be further improved, so that the small sample classification accuracy is improved.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a small sample remote sensing image scene classification method based on iterative feature distribution learning. Background technique [0002] With the development of artificial intelligence technology, deep learning can achieve satisfactory results in some scenarios by virtue of deep and complex network models, huge training data support and powerful hardware support in computer vision tasks. However, for a small number of (single) sample learning tasks, deep learning cannot improve the learning ability from complex network models and large amounts of training data. Therefore, the few-shot learning technique that learns data patterns from a few (single) samples has become a hot research direction in the current deep learning field. [0003] Small sample learning aims to achieve the ability to identify unknown categories through a small number of training s...

AI Technical Summary

Problems solved by technology

[0005] The technical problem to be solved by the present invention is to provide a small sample remote sensing image scene classification method based on iterative feature distribution learning for the problem that the deep network model is easy to overfit and the classification performance is poor when the number of samples is insufficient, and its structure is simple , reasonable design

Images