Published Paper Details:

Segmentation of water bodies from remote sensing satellite images is crucial for processing and analysis. Efficiently extracting water body features is essential for various real-time applications such as water resource allocation, ecological services assessment, monitoring systems for water resource protection, and identification of flooding disasters. The interpretation of satellite images and the random extraction of water bodies are emerging tasks in remote sensing. Convolutional Neural Networks (CNNs) have shown promise in various remote sensing applications, including efficient segmentation of water bodies from satellite images. Several CNN-based approaches have been proposed for this purpose. Multi-scale Water Extraction CNN (MWEN) stands out as a particularly effective method for segmenting water from Geo-Fan-2 (GF-2) sensing satellite images. However, addressing the automatic extraction of water body segmentation from satellite images remains challenging due to the sparse arrangement of boundary pixels and the large number of training samples required. To enhance the performance of automatic water body segmentation from satellite images, we propose a Novel Optimized Multi-Feature Contour-based Hierarchical Neural Network (NOMFCHNN). NOMFCHNN incorporates expanding neural network features and layers inspired by inception models, which provide crucial information about network localization. This approach employs pixel matching with extended feature extraction. Extensive experiments on combined satellite image repositories demonstrate that our proposed approach improves segmentation accuracy and other key parameters compared to state-of-the-art methods for remote sensing satellite images.