Shadow-Resilient Greenery Index Estimation Using Deep Learning

Ashik Mahmud Raz; Md Anamul  Islam; Azmain Inquaid Haque; Aminul Islam; G M Atiqur  Rahaman

doi:10.53808/KUS.2026.23.02.1629-se

Authors

Ashik Mahmud Raz Computer Science and Engineering Discipline, Khulna University, Khulna-9208, Bangladesh
Md Anamul Islam Computer Science and Engineering Discipline, Khulna University, Khulna-9208, Bangladesh
Azmain Inquaid Haque Computer Science and Engineering Discipline, Khulna University, Khulna-9208, Bangladesh
Aminul Islam Computer Science and Engineering Discipline, Khulna University, Khulna-9208, Bangladesh
G M Atiqur Rahaman Computer Science and Engineering Discipline, Khulna University, Khulna-9208, Bangladesh

DOI:

https://doi.org/10.53808/KUS.2026.23.02.1629-se

Keywords:

Green Area Factor (GAF), Semantic Segmentation, Urban Greenery, Satellite Imagery, UAV Imagery, Shadow Aware Learning, Deep Learning, Urban Sustainability

Abstract

The Green Area Factor (GAF) is widely used to evaluate urban ecological quality, but conventional field-based assessment is time-consuming, labor-intensive, and difficult to apply over large areas. Although image-based methods offer a more scalable alternative, accurately identifying vegetation remains challenging because shadows cast by buildings, trees, and other urban structures often obscure green surfaces and lead to segmentation errors. To address this challenge, this study proposes a shadow-resilient deep learning framework for estimating the Greenery Index (GI) from aerial imagery as a scalable proxy for GAF. The framework combines InternImage-T with UPerNet and ResNet-34 with UNet++ using a class-wise maximum-confidence ensemble strategy. Instead of averaging predictions from multiple models, each semantic class is assigned to the model that demonstrates the highest validation performance, allowing the complementary strengths of both architectures to be fully utilized while reducing shadow-induced misclassification and improving boundary localization. The proposed framework was evaluated on two independent aerial imagery datasets acquired from satellite and UAV platforms. It achieved mIoU values of 0.88 and 0.82, F1-scores of 0.93 and 0.90, and pixel accuracies of 0.97 and 0.96. Compared with a previously published weighted-ensemble method evaluated on the same benchmark dataset, the proposed approach improved mean IoU and F1-score by six and eleven percentage points, respectively. In addition, the estimated Greenery Index showed a strong correlation with field-measured GAF values (Pearson correlation coefficient, r = 0.953) while reducing the mean absolute error from 0.140 to 0.087. These findings demonstrate that improving segmentation in shadowed regions leads to more reliable ecological indicators and provides an effective, scalable solution for urban greenery assessment, environmental monitoring, and sustainable urban planning.

Downloads

Download data is not yet available.

References

Chen, L.-C., Papandreou, G., Kokkinos, I., Murphy, K., & Yuille, A. (2018). DeepLab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs. IEEE Transactions on Pattern Analysis and Machine Intelligence, 40(4), 834–848.

Chen, W., Li, J., & Zhang, Y. (2023). Hybrid CNN-Transformer network for semantic segmentation of high-resolution remote sensing images. ISPRS Journal of Photogrammetry and Remote Sensing.

Cheng, B., Xiao, B., Wang, J., Shi, H., Huang, T. S., & Zhang, L. (2021). BiSeNet V2: Bilateral network with guided aggregation for real-time semantic segmentation. International Journal of Computer Vision, 129, 3051–3068.

Diakogiannis, F. I., Waldner, F., Caccetta, P., & Wu, C. (2020). ResUNet-a: A deep learning framework for semantic segmentation of remotely sensed data. ISPRS Journal of Photogrammetry and Remote Sensing, 162, 94–114.

Dimitrovski, I., Spasev, V., Loshkovska, S., & Kitanovski, I. (2024). U-net ensemble for enhanced semantic segmentation in remote sensing imagery. Remote Sensing, 16(12), 2077.

He, J., & Ma, R. (2023). Multimodal semantic segmentation of remote sensing images with CNN-transformer fusion. Remote Sensing.

Khan, B. A., & Jung, J.-W. (2024). Semantic segmentation of aerial imagery using u-net with self-attention and separable convolutions. Applied Sciences, 14(9), 3712.

Li, R., Zhang, M., & Wang, Z. (2024). Change detection in satellite imagery using transformer-based Siamese networks. Remote Sensing.

Li, X., Zhang, Y., Ding, K., & Zhang, L. (2022). Transformer-based deep learning for remote sensing image semantic segmentation. Remote Sensing, 14(7), 1636.

Liu, H., & Chen, X. (2024). Vision transformer adapters for remote sensing semantic segmentation. IEEE Geoscience and Remote Sensing Letters.

Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., & Guo, B. (2021). Swin Transformer: Hierarchical vision transformer using shifted windows. Proceedings of the IEEE/CVF International Conference on Computer Vision, 10012–10022.

Ma, L., Liu, Y., Zhang, X., & Ye, Y. (2022). Attention-based U-Net for remote sensing image segmentation. IEEE Access, 10, 54212–54223.

Nan, G., Li, H., Du, H., Liu, Z., Wang, M., & Xu, S. (2024). A semantic segmentation method based on AS-Unet++ for power remote sensing of images. Sensors, 24(1), 269.

Nsiah, R. A., Mantey, S., & Ziggah, Y. Y. (2023). Building segmentation from UAV orthomosaics using unet-resnet-34 optimised with grey wolf optimisation algorithm. Smart Construction and Sustainable Cities, 1(1), 21.

Panboonyuen, T., & Jitkajornwanich, K. (2023). Road segmentation on aerial images using deep convolutional neural networks. IEEE Transactions on Intelligent Transportation Systems.

Panghurian, F. P., Pranoto, H., Irwansyah, E., & Pramudya, F. S. (2024). Comparison of Resnet Models in UNet Classifier for Mapping Oil Palm Plantation Area with Semantic Segmentation Approach. International Journal of Advanced Computer Science & Applications, 15(7).

Rahaman, G. A., Längkvist, M., & Loutfi, A. (2024). Deep learning based automated estimation of urban green space index from satellite image: A case study. Urban Forestry & Urban Greening, 97, 128373.

Ronneberger, O., Fischer, P., & Brox, T. (2015). U-Net: Convolutional networks for biomedical image segmentation. Medical Image Computing and Computer-Assisted Intervention (MICCAI), 234–241.

Santosa, H., & Kusuma, G. P. (2025). Urban area road scene segmentation using InternImage-adapter model. Commun. Math. Biol. Neurosci., 2025, Article-ID.

Shen, J., Huo, C., & Xiang, S. (2024). Siamese InternImage for Change Detection. Remote Sensing, 16(19), 3642.

Shi, Y., & Luo, P. (2024). Fine-grained land use classification using transformer-based semantic segmentation. IEEE Transactions on Geoscience and Remote Sensing.

Slyusar, V., Protsenko, M., Chernukha, A., Melkin, V., Petrova, O., Kravtsov, M., Velma, S., Kosenko, N., Sydorenko, O., & Sobol, M. (2021). Improving a neural network model for semantic segmentation of images of monitored objects in aerial photographs. Eastern-European Journal of Enterprise Technologies, 6(2), 114.

Wang, J., Sun, K., Cheng, T., Jiang, B., Deng, C., Zhao, Y., Liu, D., Mu, Y., Tan, M., & Wang, X. (2020). Deep high-resolution representation learning for visual recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence.

Wang, J., Sun, L., & Huang, K. (2023). Ensemble deep learning for land cover semantic segmentation from remote sensing images. IEEE Access.

Wang, W., Dai, J., Chen, Z., Huang, Z., Li, Z., Zhu, X., Hu, X., Lu, T., Lu, L., Li, H., & others. (2023). Internimage: Exploring large-scale vision foundation models with deformable convolutions. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 14408–14419.

Wang, Z., Li, J., Tan, Z., Liu, X., & Li, M. (2023). Swin-UPerNet: A semantic segmentation model for mangroves and spartina alterniflora loisel based on UPerNet. Electronics, 12(5), 1111.

Wu, Y., & Zhang, M. (2024). Swin-CFNet: An attempt at fine-grained urban green space classification using Swin transformer and convolutional neural network. IEEE Geoscience and Remote Sensing Letters, 21, 1–5.

Xiao, T., Liu, Y., Zhou, B., Jiang, Y., & Sun, J. (2018). Unified perceptual parsing for scene understanding. Proceedings of the European Conference on Computer Vision (ECCV), 418–434.

Xie, E., Wang, W., Yu, Z., Anandkumar, A., Alvarez, J. M., & Luo, P. (2021). SegFormer: Simple and efficient design for semantic segmentation with transformers. Advances in Neural Information Processing Systems.

Xu, Q., Wang, H., & Li, M. (2024). Multi-scale semantic segmentation of satellite imagery using transformer-enhanced UNet. Remote Sensing.

Yu, J., Zhang, Y., Sun, F., Wang, L., & Lu, R. (2024). Solution for CVPR 2024 UG2+ Challenge Track on All Weather Semantic Segmentation. arXiv Preprint arXiv:2406.05837.

Yuan, Q., Wei, Y., & Meng, X. (2021). Deep learning-based semantic segmentation of remote sensing images: A review. Remote Sensing, 13(23), 4772.

Zhang, L., & Du, B. (2025). A survey on semantic segmentation of remote sensing images using deep learning. IEEE Access.

Zhao, H., Shi, J., Qi, X., Wang, X., & Jia, J. (2017). Pyramid scene parsing network. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2881–2890.

Zhou, W., & Li, Q. (2022). Improved DeepLabV3+ for semantic segmentation of satellite images. Sensors, 22(14), 5312.