Academic Journal of Computing & Information Science, 2025, 8(9); doi: 10.25236/AJCIS.2025.080906.
Hongyan Fan
Institute of Cultural and Tourism, Leshan Vocational and Technical College, Leshan, 614000, China
Historical building surveys have long faced challenges of low efficiency and high costs. This study addresses the preservation needs of historical buildings in the Leshan region by proposing a satellite image recognition framework based on an improved lightweight MobileNetV2 convolutional neural network (CNN). The model adapts to the architectural and image characteristics of the Leshan region, optimizes sample quality to enhance data usability, and incorporates an attention mechanism to improve architectural feature recognition capability. Experimental results demonstrate that the model operates with low hardware requirements, achieving 90.9% accuracy on the test set. It enables rapid identification and dynamic monitoring of historical buildings across the entire Leshan region, providing technical support for cultural heritage preservation efforts in the area.
CNN, Deep Learning, Historic Buildings, Leshan Area
Hongyan Fan. Research on Historical Building Recognition in Leshan Satellite Images Using CNN. Academic Journal of Computing & Information Science (2025), Vol. 8, Issue 9: 40-45. https://doi.org/10.25236/AJCIS.2025.080906.
[1] Zhongshu Zhao, Weijie Lan. Formation and Evolution of Key Concepts in the Protection System for Historic and Cultural Cities [J]. Urban Planning, 2022, 46(S2): 20-26.
[2] Lizhong Wang, Honghai Zhang, Bo Zhong, et al. Deep Learning-Based Method for Identifying Township Buildings in High-Resolution Remote Sensing Images [J]. Information Technology and Application in Scientific Research, 2019, 10(01): 88-95.
[3] Zedi Gao, Jianfeng Gao. Building Recognition Method for Landsat Satellite TM Images Based on Convolutional Neural Networks [J]. Information Technology and Informatization, 2020, (11): 196-199.
[4] Bin Xiong, Shuangde Zhang. A Semantic Segmentation Algorithm for Buildings in Satellite Remote Sensing Images Based on an Improved PSPNet [J]. Remote Sensing Information, 2023, 38(04): 73-79. DOI: 10.20091/j.cnki.1000-3177.2023.04.009.
[5] Zehui Li ,Ningning Zhu ,Chong Zhao, et al. Research on Building Type Identification Methods Based on Deep Learning of Satellite Images: The Case of Singapore's “Shophouses” [J]. New Architecture, 2025, (02): 93-98.
[6] Chenchen Liu,Xiaosan Ge,Yongbin Wu. Building Change Detection Integrating Twin Neural Networks and Mutual Attention [J]. Remote Sensing Information, 2024, 39(05): 70-77. DOI: 10.20091/j. cnki.1000-3177.2024.05.009.
[7] Sandler M, Howard A, Zhu M, et al. MobileNetV2: Inverted residuals and linear bottlenecks [C] // 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2018.