Hyperspectral image tree species classification method and device based on generative adversarial network

Abstract

The invention discloses a hyperspectral image tree species classification method and device based on a generative adversarial network, and the method comprises the steps: carrying out the image segmentation of a hyperspectral image, and carrying out the image reconstruction; extracting texture features and spectral features of the hyperspectral image; inputting the texture features and the spectral features into a generative adversarial network based on a combination attention mechanism to obtain image features with attention, and sending the image features into a discriminator to obtain output features of the discriminator after maximization optimization; and optimizing the generation distribution of the generator according to the minimization of score values obtained by the input features, and sending the final output features of the discriminator to the classifier to obtain a classification result. According to the method, the distribution coverage condition of various tree species of the forest can be obtained, monocular tree species type classification is achieved, monitoring and supervision are facilitated, meanwhile, the number of samples needed for data preprocessing and manual collection is small, the method is suitable for most geographic areas and land cover types, and the method is high in feasibility, robustness and prediction result accuracy.

Description

technical field [0001] The invention relates to the field of data processing, in particular to a method and device for classifying hyperspectral image tree species based on a generative confrontation network. Background technique [0002] Compared with conventional field surveys, remote sensing technology can monitor forest ecosystems quickly and efficiently. An important direction of forestry remote sensing is tree species classification and identification technology. Some scholars have combined lidar and hyperspectral data to classify forest tree species and have made positive progress. However, the acquisition cost of lidar data is expensive and affected by the area of ​​​​the flight area. Outlook is limited. Multispectral satellite remote sensing is affected by spatial resolution and spectral resolution, and it also has limitations for fine classification of ground objects. Hyperspectral data can discover more vegetation information. At present, UAV hyperspectral data ...

AI Technical Summary

Problems solved by technology

[0004] However, both of the above two schemes have disadvantages: on the one hand, supervised classification requires a large amount of sample data and a lot of energy to preprocess the sample data, and the efficiency cannot achieve the expected effect; on the other hand, the spectral features extracted by unsupervised classification When applied to monocular tree species classification, problems such as high canopy density and approximate spectral features are often encountered, resulting in a large error between the model prediction value and the real value, and the classification accuracy is not ideal

Therefore, the above two types of methods have their own defects, and their robustness and applicability are not strong.

Images