Deep Learning – Based Change Detection For Identifying Illegal Encroachments On Water Bodies

The Run Inference module allows users to upload satellite images and conduct real-time segmentation of the images through the U-Net deep learning algorithm. The program offers various file formats including PNG, JPG, and GeoTIFF for users to select their preferred data format. A user can select custom images for uploading or use sample satellite images provided by the system. These sample images represent urban landscapes, wetlands, and lake areas. Once the user uploads the selected satellite image, the model analyses the image and produces a segmentation result. The segmentation result includes categorized segments and their percentage values. In the presented case, the model classifies the selected segment into the following categories: ·        Built-up area: 27% ·        Road area: 17% ·        Land area: 56% ·        Water and vegetation: 0% Interpretation Based on the provided results, the selected segment is mainly an urban landscape with relatively high percentages of built-up and land areas. There is no water or vegetation segment in the selected region. This could mean that the area has been developed or lacks water bodies. The U-Net model effectively categorizes the land-use segments, making it easier for the model to identify the encroached areas.

[1]  O. Ronneberger, P. Fischer, and T. Brox, "U-Net: Convolutional Networks for Biomedical Image Segmentation," MICCAI, 2015. doi:10.1007/978-3- 319-24574-4_28 [2]   J. Long, E. Shelhamer, and T. Darrell, "Fully Convolutional Networks for Semantic Segmentation," CVPR, 2015. doi:10.1109/CVPR.2015.7298965 [3]   L.-C. Chen, G. Papandreou, F. Schroff, and H. Adam, "Rethinking Atrous Convolution for Semantic Image Segmentation," IEEE TPAMI, 2018. doi:10.1109/TPAMI.2017.2699184 [4]   V. Badrinarayanan, A. Kendall, and R. Cipolla, "SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation," IEEE TPAMI, 2017. doi:10.1109/TPAMI.2016.2644615 [5]  I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning, MIT Press, 2016. doi:10.7551/mitpress/10243.001.0001 [6]   X. X. Zhu et al., "Deep Learning in Remote Sensing: A Review," IEEE Geoscience and Remote Sensing Magazine, 2017. doi:10.1109/MGRS.2017.2762307 [7]   S. Ji, S. Wei, and M. Lu, "Fully Convolutional Networks for Multisource Building Extraction from an Open Aerial and Satellite Imagery Dataset," IEEE TGRS, 2018. doi:10.1109/TGRS.2018.2806216