Experimental Evaluation of Learning Based River Path Sensing and Classification Using Digital Satellite Images

The river system is constantly stressed out by the demands for electricity, water, and food, despite its crucial role in the planet's long-term viability. Automatic river mapping using satellite data has tremendous potential, according to recent advancements in machine learning. For sound management and policy choices, precise and up-to-date water extent data may be provided via surface river mapping. Nevertheless, due to imperfect labeling, geographical variability, and noise in satellite data, precise large-scale river mapping continues to be a formidable challenge. The accessibility of excellent quality imagery from satellites, user-friendly programming tools, and high-end consumer computing power have all contributed to the recent alignment of the information analytics and satellite imagery groups. The area of remote sensing and image processing faces a formidable obstacle in the process of water body segmentation. The generation of index characteristics by merging distinct spectra is a common component of classical strategies for countering this problem. On the other hand, these approaches rely heavily on rules and don't account for context. This paper introduces a novel approach to river path identification and classification using the Estimated Learning for Path Sensing and Classification (ELPSC) model. To assess the effectiveness of the proposed scheme, it is cross-validated with the conventional Support Vector Machine (SVM) classification model. Even with sparse and noisy data, we show that the suggested contrasting model of learning can successfully map a number of rivers.