Hyperspectral Image Classification Based On Deep Learning

Hyperspectral image classification has been a hot research topic in the field of remote sensing.Hyperspectral remote sensing images contain a large amount of features of the spectral information and spatial information,can use the information for classification and identification features.How to extract the deep features makes the features easier to classify,which is the next research hotspot in the field of high spectral image classification.The convolutional neural network as a major model of deep learning has become an important application in image processing direction.How to build a network for specific problems is also one of the research contents of deep learning.This paper studies the classification of hyperspectral images based on deep learning,and the main work and innovation points are as follows:(1)According to the characteristics of hyperspectral data "atlas",this paper puts forward an improved based on the convolutional neural network classification of hyperspectral data spectrum information turn gray method,introduces in detail the method of gray level the hyperspectral image data pretreatment process.Based on the training and testing of the data sets of Indian Pines and Pavia University,the neural network classification model of convolution neural network with stable performance was trained by reasonable adjustment of network parameters.The experimental results show that the classification method based on the spectral information gray scale based on the convolution neural network model can achieve better classification results than other methods.(2)To take advantage of spatial characteristics of hyperspectral image,this paper proposes a model of hyperspectral image reconstruction based on convolutional neural network to enhance space characteristics,using a new band selection strategy based on spatial characteristics,through the convolutional neural network training hyperspectral data,suitable for parameter optimization of hyperspectral image reconstruction model is set up,finally to classify the image reconstruction.The experimental results show that the classification of the reconstructed images is better than that of the original hyperspectral images.