Research On Classification Method Of Hyperspectral Image Based On Deep Learning

With the progress of science and technology,the resolution of satellite remote sensing and aerial remote sensing images has been continuously improved.The research on high-resolution satellite remote sensing images has received more and more attention from researchers in various countries.Ground classification of remote sensing images has always been a hot research area.Due to the introduction of deep learning,it has brought new breakthroughs in the research of remote sensing images.The specific content is as follows:The characteristics of hyperspectral remote sensing images are analyzed.The classification strategies of hyperspectral remote sensing images are studied and discussed.The feature selection and feature extraction of hyperspectral remote sensing images are studied.A hyperspectral image classification method based on deep neural network is introduced.The criteria for classification accuracy evaluation of hyperspectral remote sensing images,such as confusion matrix,overall accuracy,average accuracy and Kappa coefficient,are described.The purpose of land cover classification is to classify each pixel.Dividing remote sensing images into different types of land cover can be regarded as multi-level semantic processing tasks.In the thesis,a method based on convolutional neural network(CNN)for automatic classification of hyperspectral remote sensing images is proposed.The hyperspectral image is reduced by principal component analysis(PCA)technology,and the traditional CNN framework is optimized.The Inception structure is added,and the actual classification effect between it and the support vector machine(SVM)algorithm is compared horizontally.The classification accuracy of the proposed method is significantly improved,and the texture features are more prominent.