A Remote Sensing Image Classification Method Based on Multimodal Deep Learning

Abstract

The invention discloses a remote sensing image classification method based on multi-modal deep learning. The method comprises the following steps: first constructing a multi-modal remote sensing image data sample set, including multi-modal remote sensing images with different imaging principles; based on different data Modality, build a data modality-sensitive feature learning deep network, use different modal data to train the corresponding feature learning deep network; establish a feature association model between modalities to generate associated shared features and train; use the test sample set to input training fine-tuning The multi-modal deep network realizes accurate classification of remote sensing images. Its advantages are as follows: effective use of multi-source remote sensing images, mining complementary and cooperative information between modalities, the whole process is carried out automatically, which reduces the low classification accuracy caused by manual input and subjective human factors, and improves the classification accuracy.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a remote sensing image classification method based on multimodal deep learning. Background technique [0002] Classification is an important basis for the application of remote sensing image analysis. Based on the multi-platform, multi-spectrum, and multi-channel continuous observation of the ocean, the "star-machine-earth" three-dimensional observation network has spawned multi-scale, multi-temporal, multi-azimuth, and multi-level ocean remote sensing images, providing a richer and more accurate description of ground objects. data information. There is information complementarity and cooperation among multi-source remote sensing images. Using two or more remote sensing data sources to extract information can obtain higher extraction accuracy than using any one of the remote sensing data sources alone. Therefore, the abundant and available multi-source remote sensing i...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to address the deficiencies in the prior art, to provide a remote sensing image classification method based on multimodal deep learning, to solve the problem of low classification accuracy caused by manual extraction of low-level features and the limitation of information contained in single-modal data. Learn to extract the high-level features of remote sensing images step by step, and effectively associate the complementary, cooperative and redundant information between multi-source remote sensing images, and then obtain high-level associated shared features to achieve accurate classification of remote sensing images

Images