Deep learning model suitable for small sample hyperspectral image classification

Abstract

The invention provides a deep learning model suitable for small sample hyperspectral image classification, and the overall framework of the model is based on a coder-decoder, and the deep learning model comprises the steps: inputting a to-be-classified hyperspectral image and a data set; randomly sampling the original data set twice to generate a group of random training sample sequences; extracting spectral dependence features in a long-short range and a spatial relationship of pixels in a local range by utilizing a global convolution long-short-term memory module; respectively extracting detailed spectral dependence features and spatial details by utilizing a global spectrum and a spatial attention mechanism; and recovering space details of the semantic features by utilizing a feature migration module. The invention provides a novel global learning classification method, which not only can fully excavate the dependence between spectral features in a long range and a short range, butalso can extract the dependence between pixel spatial features in the long range and the short range, so that the most discriminant feature can still be extracted when a training sample is limited, and the classification precision is ensured.

Description

technical field [0001] The invention relates to the combination of deep learning and remote sensing image processing fields, mainly solving the problem of classification of remote sensing image features, and specifically relates to a deep learning model suitable for small-sample hyperspectral image classification. Background technique [0002] With the rapid development of remote sensing technology, a large amount of remote sensing image data with high spatial resolution and high spectral resolution is becoming easier to obtain. Therefore, how to extract valuable information from these remote sensing images with rich spectral and spatial information has always been one of the hotspots in academic research. Among them, hyperspectral image classification is one of the most important applications. Hyperspectral image data has the characteristics of high dimensionality and huge data volume. Because of this, hyperspectral imagery is widely used in target detection, agricultural ...

AI Technical Summary

Problems solved by technology

Although these methods have better classification results, these methods require manual selection of features and cannot achieve end-to-end training.

Images