Optical remote sensing scene classification method and apparatus based on deep twin residual networks

Abstract

The invention discloses an optical remote sensing scene classification method based on deep twin residual networks, and belongs to the technical field of image processing. In a training phase, featureextraction is performed on input remote sensing image pairs via two completely the same depth residual networks to obtain feature representations of the input remote sensing image pairs respectively,and then an Euclidean distance between the input remote sensing image pairs in a feature space is calculated by combining the two feature representations to judge the similarity between the input image pairs. In a test phase, scene classification is performed on the input images by using any one of the trained deep residual networks. By adoption of the optical remote sensing scene classificationmethod disclosed by the invention, the scene classification can be performed on large-scale high-resolution remote sensing images, and the optical remote sensing scene classification method can play an important role in natural disaster monitoring and evaluation, urban planning, environmental monitoring and other fields.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to an optical remote sensing image scene classification method based on a deep twin residual network in the technical field of remote sensing image processing. Background technique [0002] With the development and extension of deep learning, deep learning networks have made some progress in various fields, and the field of remote sensing is no exception. In recent years, the rapid development of aviation and aerospace remote sensing technology is even more powerful. As an important application in the field of remote sensing, remote sensing image classification has attracted the attention of relevant professionals, and more and more energy has been devoted to it. The methods of remote sensing image classification are mainly divided into two categories, one is the classification method using non-deep learning, and the other is the classification method combined with d...

AI Technical Summary

Problems solved by technology

[0006] (2) The data lacks diversity: due to the overall small scale of the data, the image information provided by the data set is naturally not rich enough

[0007] Problems in remote sensing datasets greatly limit the development of deep learning networks in scene classification applications

The limited data size makes the feature expression learned by these networks not robust, and the network is prone to overfitting

Images