Hyperspectral image classification method based on binary quantization network

Abstract

The invention provides a hyperspectral image classification method based on a binary quantization network. The method mainly solves the technical problems that edges of hyperspectral images are fuzzy,models are complex and spatial feature information of extracted clouds is insufficient. The method comprises the steps of generating a training set; constructing a full-precision convolutional neuralnetwork; training a full-precision convolutional neural network; constructing a binary quantization convolutional neural network; initializing a binary quantization convolutional neural network; training the binary quantization convolutional neural network; and classifying the hyperspectral images. According to the hyperspectral image cloud classification method, seven quantization modules are utilized, a traditional convolutional neural network structure is simplified, spatial feature information of a hyperspectral image cloud layer is fully extracted, and the precision of hyperspectral image cloud classification is improved while computing resources are saved.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a hyperspectral image classification method based on a binary quantization network in the technical field of image classification. The invention can be used to classify cloudy images and cloudless images from hyperspectral images. Background technique [0002] Hyperspectral imagers can image objects in hundreds of narrow bands, ranging from visible light to infrared bands, and the feature of "integration of images and spectra" enables them to obtain more information. At present, hyperspectral data processing has become an important research field at home and abroad. Among them, hyperspectral image cloud classification has important theoretical value and application prospects in aviation flight support and weather forecasting. According to the proportion of clouds, hyperspectral images can be divided into two categories: less cloudy and more cloudy. However, hype...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to address the deficiencies in the prior art above, and propose a hyperspectral image classification method based on a binary quantization network, which is used to solve the image edge blurring that exists when the existing hyperspectral image classification method is used for cloud classification , the technical problems of high method complexity, large amount of computation and low classification accuracy

Images