Research On Remote Sensing Image Classification Based On Deep Convolutional Neural Network

With the rapid development of modern remote sensing technology,the resolution of remote sensing images is getting higher and higher,and the information of ground objects in remote sensing image is becoming more and more abundant,which has become one of the powerful tools for people to understand the earth and the surrounding environment.How to effectively identify the information in remote sensing images is one of the important contents of remote sensing image analysis.Compared with traditional methods,deep convolutional neural networks can automatically,efficiently and accurately identify objects in remote sensing image.The trained convolutional neural network model can easily and automatically extract features from the image.The essence of convolutional neural networks is to fit the mapping relationship from input to output,which includes two basic steps:feature extraction and feature mapping.Moreover it can reduce the training difficulty through local receptive field and weight sharing.Remote sensing image classification based on deep learning model can fully extract the semantic features contained in the image,and the semantic information in the image can improve the classification accuracy of the remote sensing image.Based on this,this paper combines the Residual neural network(ResNet)with the Feature pyramid networks(FPN)algorithm to construct a suitable neural network model to study the classification of images.The main work of the paper is as follows:(1)Build a deep learning environment,and preprocess GID data according to the deep learning framework and experimental environment.Then,train ResNet with preprocessed dataset,and appropriate hyper-parameters are set in the process,to make the model converge after training.To prevent overfitting,Dropout is introduced.Finally,the experimental results are analyzed and compared with the results of support vector machine(SVM)that a traditional remote sensing classification method.By comparing the two classification results,it is found that the accuracy of remote sensing classification based on residual neural network is significantly better than SVM method.(2)To further improve the accuracy of remote sensing image classification,the ResNet-FPN high-resolution image remote sensing classification model is constructed based on ResNet and FPN feature fusion algorithm.Using the advantage of FPN,the multi-layer semantic features extracted from images can be fused to improve the utilization ratio of features.The experimental results show that the improved model has improved the classification accuracy significantly compared with the residual neural network,and has a good effect on edge segmentation and improves the recognition rate of small objects in remote sensing images.