Improved deep convolutional neural network-based remote sensing image classification model

Abstract

The invention relates to an improved deep convolutional neural network-based remote sensing image classification model. The model comprises the following steps of: S1, carrying out dimensionality reduction on a remote sensing feature image on the basis of a bottleneck unit; S2, carrying out convolutional multichannel optimization on the remote sensing feature image on the basis of grouped convolution; S3, improving feature extraction ability of the remote sensing feature image on the basis of channel shuffling; and S4, carrying out band processing on spatial position features of the remote sensing image. The model has the advantages that the dimensionality reduction of to-be-input remote sensing images is realized, and the convolutional calculation amount during the training of deep convolutional neural network-based remote sensing image classification model is reduced; a channel shuffling structure is constructed in allusion to spatial correlation of the remote sensing images, so thatthe feature extraction ability of a neural network in the grouped convolution stage is enhanced; and aiming at spatial position features of the remote sensing images, spatial position feature recognition degrees, for the remote sensing images, of the deep convolutional neural network-based model are improved.

Description

technical field [0001] The invention relates to the technical field of remote sensing images, in particular to a remote sensing image classification model based on an improved deep convolutional neural network. Background technique [0002] In the early 1970s, the United States took the lead in launching the Earth Resources Satellite Program and began to conduct long-distance and non-contact observations of the earth's surface. Driven by it, the multi-platform-based earth observation technology has developed rapidly worldwide. my country's "National Medium and Long-term Science and Technology Development Plan (2006-2020)" clearly points out that it is necessary to build an advanced earth observation system with multi-spectral segments, different orbital heights, and all-weather, all-weather, and global observation capabilities, and form an independent intellectual property rights system. spatial information industry chain. The rapid development of earth observation technolog...

AI Technical Summary

Problems solved by technology

However, there is no corresponding disclosure about a technical solution that can reduce the consumption of computing resources, ensure the effect of feature extraction, and improve the recognition degree of spatial position features.

[0010] In summary, there is a need for a remote sensing image classification model based on an improved deep convolutional neural network that can reduce the consumption of computing resources, ensure the effect of feature extraction, and improve the recognition of spatial location features. See the report

Images