Classification Algorithm Based On Deep Learning For Remote Sensing Image

Remote sensing image classification has been widely used in fields such as military reconnaissance,urban planning,and environmental monitoring,and has received increasing attention.With the development of deep learning,remote sensing image classification based on deep learning has attracted widespread attention from researchers and has made great progress.However,there are still many challenges in current remote sensing image classification based on deep learning,such as scarce labeled datasets used for training,complex image backgrounds,and large scale variations.Based on the characteristics of remote sensing images,this paper conducts in-depth research on remote sensing image classification tasks using deep learning techniques,and achieves the following research results:(1)Proposed a remote sensing image classification model based on transfer learning and channel attention.In response to the problems of a small number of training samples for remote sensing images,complex backgrounds,and redundant geographic features,channel attention was added to the convolutional neural network and a transfer learning training method was adopted to propose a remote sensing image classification model based on transfer learning and channel attention.This model first uses two convolutional neural networks as the backbone,and adds channel attention mechanism to adaptively enhance the main features and suppress the minor features;then,the features extracted by these two networks are fused together,and finally fine-tuning transfer learning is used to achieve learning and classification in the target domain.After testing and verifying on multiple public datasets,experimental results prove that the method proposed in this article has great performance in remote sensing image classification tasks,achieving performance comparable to other advanced methods.(2)A CNN-based remote sensing image classification model called CVi T is proposed.To address the issues of poor performance of vision Transformers with limited training samples and weak local feature extraction ability,this paper first uses convolutional modules to extract local features from the input remote sensing image.Then,the extracted features are classified using Vision Transformers,and finally,fine-tuning transfer learning is adopted for remote sensing image classification.After classification accuracy testing on three public datasets,experimental results demonstrate that the CVi T model achieves an approximately 0.9% to 1.1% increase in classification accuracy compared to convolutional neural networks and original Vision Transformers.