Hyperspectral image classification method based on deep feature cross fusion

Abstract

The present invention proposes a hyperspectral image classification method based on deep feature cross-fusion, which mainly solves the problem of loss of spatial features in traditional convolutional neural network classification of hyperspectral data. Its technical solution is: 1. Read hyperspectral data and preprocess each spectral band; 2. Use the preprocessed hyperspectral data to construct data samples and generate training set and test set data; 3. Build depth-based features Cross-fused hyperspectral image classification network; 4. Use the training set data to train the network; 5. Use the trained network to classify and predict the test set data; the present invention aims at the multi-channel raw data fusion of different branch stages and different scales The depth features of the hyperspectral data are continuously exchanged between multi-scale representations, thereby improving the deep feature expression ability of the model; the multi-scale spatial information of the depth features of different layers of hyperspectral data is effectively used, and the classification accuracy is improved.

Description

technical field [0001] The invention belongs to the technical field of remote sensing information processing, and further relates to a hyperspectral image classification method, specifically a hyperspectral image classification method based on depth feature cross fusion, which can realize ground object recognition and is used in the fields of environmental monitoring, geological exploration and the like. Background technique [0002] With the development of spectral imaging technology, the spatial resolution of hyperspectral images has been continuously improved, and there are more and more spectral bands, making the information of hyperspectral images more and more abundant. Rich spectral and spatial features make hyperspectral image classification more promising, and at the same time, the classification accuracy requirements are more stringent. [0003] Hyperspectral image classification technology mainly includes data feature engineering and classification. Feature engine...

AI Technical Summary

Problems solved by technology

The disadvantage of this method is that it does not consider the strong correlation between deep features, and as the depth of the network increases, the serial convolutional neural network will extract features from high spatial resolution to low spatial resolution, resulting in spatial features. loss of

Images