Swapnil Kannojia
Computer science
Hometown: Lucknow, Uttar Pradesh, India
Graduation date: Spring 2025
MORE | Spring 2025
Global Crop Field Delineation with SAM-Based Segmentation and LSTM-Based Pixel Classification
Field boundary delineation is essential for agricultural monitoring and precision farming. Traditional methods rely on region-specific training, limiting their applicability. This study presents a globally transferable method using multi-temporal Planet.com Visual Basemaps at 3m resolution. An adaptive weighted composite image enhances field boundaries, achieving a mean monthly coverage of 94.28%, while the composite image attains 95.23% coverage, demonstrating its effectiveness. SAM2 segments boundaries using zero-shot learning. The LSTM classifier processes pixel-level temporal information to classify each segmented region as either a crop field or a non-field area, achieving 84% accuracy and outperforming traditional methods by 8-12%. Future work includes integrating high-resolution spectral bands and real-time remote sensing for improved monitoring.
Mentor: Nakul Gopalan
It’s hip to be square.
Students presenting projects at the Fulton Forge Student Research Expo are encouraged to download this personal QR code and include it within your poster. This allows expo attendees to explore more about your project and about you in the future.
Right click the image to save it to your computer.