Oil spill hyperspectral image detection method based on two-way graph u-net convolutional network

Abstract

The invention discloses an oil spill hyperspectral image detection method based on a dual-path graph U-NET convolution network, which includes the following steps: Step 1. Perform graph structuring on the data of hyperspectral images that need to detect oil spills, and obtain the spectrum. Graph structure information and spatial graph structure information; Step 2. Send the spectral graph structure information and spatial graph structure information to one channel of the dual-channel graph U-NET convolution network to obtain the spectral map of the hyperspectral image respectively. Features and spatial map features; Step 3. Fusion of the spectral map features and the spatial map features to obtain the spatial-spectral map features; Step 4. Send the spatial-spectral map features to the classifier to obtain the spatial map features. The classification results of hyperspectral images are described. The present invention can map Euclidean image data into non-Euclidean data, more effectively represent spectral information and spatial information, further extract the spatial spectral characteristics of oil spill hyperspectral images, and improve the detection accuracy of oil spill hyperspectral images.

Description

Technical field [0001] The present invention relates to the technical field of image processing, specifically to the technical field of machine learning and hyperspectral image classification, and more specifically to a hyperspectral image detection method based on a dual graph (Dual Graph) U-NET convolution network, which can be Detection of oil spill hyperspectral images. Background technique [0002] Hyperspectral imaging refers to using an imaging spectrometer to record the spectral characteristics of various ground objects, imaging each band separately, and forming a data cube from the images in different bands as a hyperspectral image. Hyperspectral images add spectral information based on two-dimensional spatial information, and each pixel has a corresponding reflection spectrum curve. Due to its continuous spectral resolution and ability to identify material surface characteristics, hyperspectral remote sensing has the ability to effectively locate and distinguish s...

AI Technical Summary

Problems solved by technology

However, the local information of high spatial resolution generated in the process of graph convolution is not fully utilized

Images