Volume 18, No. 6, 2021
Delineation Of Agricultural Field Boundaries Through A Cascaded Deep Network Model From Polarized SAR & Multispectral Images
Prof. Salunke Shrikant Dadasaheb, Prof. Bere Sachin Sukhadeo, Prof. Bhange Rajani Rahul , Prof. Kadam Swati Amol, Prof. Shende Sachin Santosh
Abstract
By combining multispectral imagery from Sentinel-1 and Sentinel-2 satellites with polarized Synthetic Aperture Radar (SAR), this research presents a cascaded deep network model for agricultural field border demarcation. The model uses a cutting-edge method that combines edge detection and semantic segmentation to improve the accuracy of recognizing agricultural boundaries. TensorFlow and PyTorch are used to create models in this study, which covers a wide range of geographical and climatic zones. Classification reports and Intersection-over-Union. (IoU) scores are examples of evaluation metrics. The results demonstrate the usefulness of the model by giving accurate field boundaries that may be used in precision farming. Suggestions include integrating topographical data for improved accuracy, improving model design, and enriching datasets even further.
Pages: 9617-9624
Keywords: Sentinel-1, Sentinel-2, IoU score, deep learning, multispectral pictures, agricultural borders, SAR remote sensing, precision agriculture, and dataset enrichment.