Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{207272,
        author = {Yogesh S. Sirsath and Dr.Ashwini S.Gaikwad and Dr.Ashwini S.Gaikwad},
        title = {An Integrated Machine-Learning and GIS Framework for Agricultural Field Delineation Using Multispectral Remote-Sensing Imagery},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {3},
        pages = {56-68},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=207272},
        abstract = {Accurate agricultural field delineation is fundamental to precision farming, crop monitoring, land administration, and GIS-based decision support. This paper presents the implementation and experimental evaluation of an integrated remote-sensing framework that combines multispectral preprocessing, vegetation-index computation, engineered spectral-textural features, supervised machine-learning classification, preliminary boundary extraction, and GIS-oriented visualization. The implementation accepts CSV feature tables and HDF5 multispectral tensors, computes NDVI, EVI, SAVI, and NDWI, derives band statistics, texture contrast, edge density, and area-related descriptors, and compares seven classifiers. Experiments conducted on ZueriCrop-derived and custom agricultural datasets show that Random Forest achieved the best overall performance, with 97% accuracy, 96% precision, 96% recall, 96% F1-score, and Cohen’s kappa of 0.95. Gradient Boosting and SVM obtained accuracies of 95% and 93%, respectively. Feature-importance analysis identified NDVI, NIR, texture contrast, edge density, SAVI, SWIR1, and field area as influential predictors. The system additionally generated vegetation maps, moisture maps, connected agricultural regions, boundary overlays, class-wise reports, and exportable GIS summaries through an interactive Streamlit interface. The findings demonstrate that combining spectral indices with statistical and textural descriptors provides a reliable and computationally practical foundation for agricultural-region classification and field-boundary analysis. The framework is modular, reproducible, and extensible toward deep semantic segmentation, multi-temporal monitoring, and automatic polygon generation.},
        keywords = {agricultural field delineation, multispectral imagery, NDVI, Random Forest, remote sensing, machine learning, GIS, precision agriculture.},
        month = {August},
        }