Next-Generation AI Vegetation Analytics: Low-Cost PSRI Translation from RGB and NDVI for Precision Crop Monitoring

Jacob Serafin, Yuvraj Gill, Lokesh Muthiah, Dylan Lewis, Gurjit Randhawa, Aitazaz Farooque
Proceedings of the The 39th Canadian Conference on Artificial Intelligence, PMLR 318:1191-1196, 2026.

Abstract

Plant senescence monitoring is important for crop health assessment and precision agriculture, but senescence-sensitive indices such as the Plant Senescence Reflectance Index (PSRI) are not always readily available in low-cost imaging workflows. This study presents an inexpensive artificial intelligence approach to translate RGB and Normalized Difference Vegetation Index (NDVI) imagery into PSRI maps using supervised image-to-image regression. Using tiled samples from the open-sourced Canadian Cropland Dataset, three models (UNet, pix2pix GAN, and R2AttUNet) and a baseline model (linear regressor) were trained and evaluated under a consistent pipeline using mean absolute error (MAE), peak signal-to-noise ratio (PSNR), and structural similarity index measure (SSIM). The three models all outperformed the baseline across all scores; of those, UNet achieved the best performance on both validation and test sets, producing the lowest MAE (0.0248) and the highest PSNR (19.85) and SSIM (0.9009) on the test set, while GAN showed competitive but weaker results and R2AttUNet underperformed. The close validation and test agreement indicates stable generalization under the current split. These results demonstrate the feasibility of estimating PSRI from inexpensive RGB+NDVI inputs and support the use of lightweight convolutional models for low-cost vegetation monitoring.

Cite this Paper

BibTeX
@InProceedings{pmlr-v318-serafin26a, title = {Next-Generation AI Vegetation Analytics: Low-Cost PSRI Translation from RGB and NDVI for Precision Crop Monitoring}, author = {Serafin, Jacob and Gill, Yuvraj and Muthiah, Lokesh and Lewis, Dylan and Randhawa, Gurjit and Farooque, Aitazaz}, booktitle = {Proceedings of the The 39th Canadian Conference on Artificial Intelligence}, pages = {1191--1196}, year = {2026}, editor = {Bouzar-Benlabiod, Lydia and Leung, Carson}, volume = {318}, series = {Proceedings of Machine Learning Research}, month = {25--29 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v318/main/assets/serafin26a/serafin26a.pdf}, url = {https://proceedings.mlr.press/v318/serafin26a.html}, abstract = {Plant senescence monitoring is important for crop health assessment and precision agriculture, but senescence-sensitive indices such as the Plant Senescence Reflectance Index (PSRI) are not always readily available in low-cost imaging workflows. This study presents an inexpensive artificial intelligence approach to translate RGB and Normalized Difference Vegetation Index (NDVI) imagery into PSRI maps using supervised image-to-image regression. Using tiled samples from the open-sourced Canadian Cropland Dataset, three models (UNet, pix2pix GAN, and R2AttUNet) and a baseline model (linear regressor) were trained and evaluated under a consistent pipeline using mean absolute error (MAE), peak signal-to-noise ratio (PSNR), and structural similarity index measure (SSIM). The three models all outperformed the baseline across all scores; of those, UNet achieved the best performance on both validation and test sets, producing the lowest MAE (0.0248) and the highest PSNR (19.85) and SSIM (0.9009) on the test set, while GAN showed competitive but weaker results and R2AttUNet underperformed. The close validation and test agreement indicates stable generalization under the current split. These results demonstrate the feasibility of estimating PSRI from inexpensive RGB+NDVI inputs and support the use of lightweight convolutional models for low-cost vegetation monitoring.} }
Endnote
%0 Conference Paper %T Next-Generation AI Vegetation Analytics: Low-Cost PSRI Translation from RGB and NDVI for Precision Crop Monitoring %A Jacob Serafin %A Yuvraj Gill %A Lokesh Muthiah %A Dylan Lewis %A Gurjit Randhawa %A Aitazaz Farooque %B Proceedings of the The 39th Canadian Conference on Artificial Intelligence %C Proceedings of Machine Learning Research %D 2026 %E Lydia Bouzar-Benlabiod %E Carson Leung %F pmlr-v318-serafin26a %I PMLR %P 1191--1196 %U https://proceedings.mlr.press/v318/serafin26a.html %V 318 %X Plant senescence monitoring is important for crop health assessment and precision agriculture, but senescence-sensitive indices such as the Plant Senescence Reflectance Index (PSRI) are not always readily available in low-cost imaging workflows. This study presents an inexpensive artificial intelligence approach to translate RGB and Normalized Difference Vegetation Index (NDVI) imagery into PSRI maps using supervised image-to-image regression. Using tiled samples from the open-sourced Canadian Cropland Dataset, three models (UNet, pix2pix GAN, and R2AttUNet) and a baseline model (linear regressor) were trained and evaluated under a consistent pipeline using mean absolute error (MAE), peak signal-to-noise ratio (PSNR), and structural similarity index measure (SSIM). The three models all outperformed the baseline across all scores; of those, UNet achieved the best performance on both validation and test sets, producing the lowest MAE (0.0248) and the highest PSNR (19.85) and SSIM (0.9009) on the test set, while GAN showed competitive but weaker results and R2AttUNet underperformed. The close validation and test agreement indicates stable generalization under the current split. These results demonstrate the feasibility of estimating PSRI from inexpensive RGB+NDVI inputs and support the use of lightweight convolutional models for low-cost vegetation monitoring.
APA
Serafin, J., Gill, Y., Muthiah, L., Lewis, D., Randhawa, G. & Farooque, A.. (2026). Next-Generation AI Vegetation Analytics: Low-Cost PSRI Translation from RGB and NDVI for Precision Crop Monitoring. Proceedings of the The 39th Canadian Conference on Artificial Intelligence, in Proceedings of Machine Learning Research 318:1191-1196 Available from https://proceedings.mlr.press/v318/serafin26a.html.

Related Material