AUTHOR=Lin Xinhao , Hei Junmiao , Wang Yixiao , Zhang Ang 

TITLE=Research on intelligent classification of coastal land cover by integrating remote sensing images and deep learning

JOURNAL=Frontiers in Environmental Science

VOLUME=Volume 13 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/environmental-science/articles/10.3389/fenvs.2025.1612446

DOI=10.3389/fenvs.2025.1612446

ISSN=2296-665X

ABSTRACT=IntroductionThe intelligent classification of coastal land cover is an essential task for effective coastal management and environmental monitoring. With the increasing availability of remote sensing images, leveraging advanced machine learning methods, such as deep learning, has become pivotal in improving classification accuracy. Traditional methods, like pixel-based and object-oriented classification, often struggle with high complexity and inaccurate results due to limitations in handling spatial relationships and spectral data.MethodsThis research addresses these shortcomings by integrating deep learning models, particularly convolutional neural networks (CNNs) and spatially dependent learning techniques, to develop a robust classification model for coastal land cover using remote sensing data. Our approach incorporates multi-scale spatial analysis and graph-based models to capture spatial dependencies and contextual features across various coastal environments. The model also emphasizes spatial continuity, enabling a more realistic representation of complex land cover types such as wetlands, beaches, mangroves, and urbanized coastlines.ResultsCompared to traditional machine learning baselines, our method achieves improvements of +10–15% in overall accuracy and +12–14% in macro F1-score, highlighting the practical advantages of deep learning in capturing spatial structures and heterogeneity. The proposed method achieves classification accuracies of 95.83% on the Gaofen image dataset and 94.34% on the LandCoverNet dataset, with F1 scores of 91.65% and 92.42% respectively.DiscussionThese results demonstrate significant improvements in both precision and robustness when applied to high-resolution coastal remote sensing images. This work highlights the potential of deep learning in enhancing remote sensing analysis for environmental and urban applications, paving the way for intelligent decision-making in dynamic coastal zones.