Abstract
Deep learning models, especially the convolutional neural networks (CNNs), are very active in hyperspectral remote sensing image classification. In order to better apply the CNN model to hyperspectral classification, we propose a CNN model based on Fletcher–Reeves algorithm (F–R CNN), which uses the Fletcher–Reeves (F–R) algorithm for gradient updating to optimize the convergence performance of the model in classification. In view of the fact that there are fewer optional training samples in practical applications, we further propose a method of increasing the number of samples by adding a certain degree of perturbed samples, which can also test the anti-interference ability of classification methods. Furthermore, we analyze the anti-interference and convergence performance of the proposed model in terms of different training sample data sets, different batch training sample numbers and iteration time. In this paper, we describe the experimental process in detail and comprehensively evaluate the proposed model based on the classification of CHRIS hyperspectral imagery covering coastal wetlands, and further evaluate it on a commonly used hyperspectral image benchmark dataset. The experimental results show that the accuracy of the two models after increasing training samples and adjusting the number of batch training samples is improved. When the number of batch training samples is continuously increased to 350, the classification accuracy of the proposed method can still be maintained above 80.7%, which is 2.9% higher than the traditional one. And its time consumption is less than that of the traditional one while ensuring classification accuracy. It can be concluded that the proposed method has anti-interference ability and outperforms the traditional CNN in terms of batch computing adaptability and convergence speed.
Highlights
Remote sensing image classification is important for environmental monitoring
Some methods were developed for the classification of hyperspectral remote sensing images, such as the methods based on sparse representation [7], metric learning [8], stacked autoencoder (SAE) [9,10,11], stacked sparse autoencoder (SSAE) [12], etc., showing the effectiveness of classification
Compared with the support vector machine (SVM) [14,15,16], boosting [17,18], maximum entropy method [19], etc. the shallow machine learning model with a hidden layer node or no hidden layer nodes, the deep nonlinear network structure of deep learning can learn the deep connection of data and effectively improve the classification accuracy
Summary
Remote sensing image classification is important for environmental monitoring. The emergence of hyperspectral imaging technology and hyperspectral remote sensing imagery provides more possibilities for remote sensing image classification. As one of the deep learning expert models, convolutional neural network (CNN) proved its successful application in remote sensing image classification [20,21,22,23,24]. The general idea is to extract the spectral, spatial, or spectral–spatial jointly from hyperspectral images into the CNN model by dimension reduction [29], filtering [30], optimal band combination [31], etc., or enter the above resulting depth features to different classifiers such as logistic regression (LR) [29,32], extreme learning machine (ELM) [31] or SVM classifiers [30]. Due to the spectral and spatial richness of hyperspectral, many experts carried out related research on 3D CNN [36,37], and proposed the 3D CNN model in hyperspectral image classification, such as end-to-end spectral–spatial residual network (SSRN) [38], superpixel-based 3D deep neural networks [39], etc.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
