Remote Sensing Image Scene Classification Based on Convolutional Neural Networks

作     者：Liu, Yumei 

作者机构：Chongqing City Vocational College Chongqing402160 China 

出 版 物：《Informatica (Slovenia)》 (Informatica)

年 卷 期：2025年第49卷第9期

页      面：45-54页

核心收录：

学科分类：08[工学] 0803[工学-光学工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：Supported by the university-level platform of Digital Creative Design Production-Education Integration Collaborative Innovation Center 

主　　题：Convolutional neural networks 

摘      要：Remote sensing image classification, a specific application of digital signal technology in remote sensing, addresses the challenge of effectively processing and categorizing remote sensing imagery. This research proposes a neural network (CNN)-based approach for remote sensing image classification, aiming to overcome the limitation of single-feature inadequacy. The method involves making a multi-site and multi-combination strategy that effectively combines spectral features, spatial patterns, and more remote sensing images as vectors or matrices. We then train the CNN model based on the length of the data. Experimental results demonstrate a significant reduction of approximately 80% in the training time for the PCA-free CNN (SST) method after implementing the PCA transformation. This reduction not only expedites the training process but also enhances overall accuracy by approximately 3.49. The CNN-style network model contributes to efficiency improvement. Larger training models increase the number of models to be taught, slowing down the training process and prolonging learning times. The incorporation of multi-location and multi-combination strategies accelerates tracking speed and enhances the classification accuracy of remote sensing images. Comparative analysis indicates that, in contrast to other classification methods, CNN achieves superior classification performance, demonstrating its capability for increased categorization and improved accuracy. © 2025 Slovene Society Informatika. All rights reserved.