Article Highlight | 26-Dec-2023

Drone-based phenotyping reduces food waste and boosts farmer profits

Plant Phenomics

In modern agriculture, the waste of nonstandard vegetables, particularly broccoli, represents a significant portion of food loss, with aesthetics-driven criteria leading to vast quantities of unharvested produce. Recent research focuses on predicting optimal harvest dates to minimize waste, incorporating factors like temperature and broccoli head size variation. However, these models often overlook individual growth differences, necessitating advancements in precision agriculture. Smart farming technologies, including remote sensing and AI, promise to address this by accurately measuring individual vegetable sizes. The primary challenges include refining photogrammetry techniques, reducing the high labor costs of image analysis, and managing the computational demands of high-throughput phenotyping. By overcoming these hurdles, smart farming has the potential to revolutionize food waste reduction in agriculture.

In September 2023, Plant Phenomics published a research article entitled by “Drone-Based Harvest Data Prediction Can Reduce On-Farm Food Loss and Improve Farmer Income”.

In this study, researchers aimed to improve the accuracy of existing temperature-based broccoli size prediction model utilizing drone-based phenotyping, offering a precise and labor-efficient pipeline for broccoli head size estimation and harvest date prediction. The method leverages an open-source Python package and focuses on organ-level analysis using RGB sensors for broader applicability and ease of use. Results demonstrated that  the drone-based pipeline could effectively detect broccoli positions with minimal training data, requiring only brief manual postprocessing. The BiSeNet model improved broccoli head segmentation through iterative training and validation, achieving high Mean Intersection over Union (Mid IoU) scores, indicative of accurate segmentation. The head diameter (HD) calculations validated through field comparisons showed a high correlation and acceptable error margins, highlighting  the method's advantages in efficiency and accuracy compared to traditional ground surveys. Furthermore, the study developed temperature-based growth models tailored for individual farms, demonstrating the feasibility of predicting head sizes and optimal harvest dates with high accuracy. Income estimations revealed the critical impact of precise harvest timing on profitability, with mere one-day deviations from the optimal date significantly affecting income. While the study successfully addressed several technical challenges, it also recognized limitations such as the need for manual inspection and the unresolved issue of leaf occlusion. The significant initial investment for the necessary hardware and software was acknowledged, yet offset by the potential substantial income gains for larger farms.

In conclusion, this research represents a substantial advancement in digital agriculture, demonstrating an actionable and efficient system for optimizing harvests and reducing waste. While primarily focusing on broccoli, the framework's potential applicability to other vegetables holds promise for broader agricultural benefits. The study's success in integrating aerial phenotyping with machine learning and deep learning techniques paves the way for sustainable, profitable, and data-driven farming practices.

###

References

Authors

Haozhou  Wang1, Tang  Li1, Erika  Nishida1, Yoichiro  Kato1, Yuya  Fukano2*, and Wei  Guo1*

Affiliations

1Graduate  School  of  Agricultural  and  Life  Sciences,  The  University  of  Tokyo,  Tokyo,  Japan.  

2Graduate School of Horticulture, Chiba University, Chiba, Japan.

About Yuya Fukano & Wei Guo

Yuya Fukano: Dr. Fukano is an associate professor at the Graduate School of Agricultural and Life Sciences, the University of Tokyo. He specializes in ecology and crop science. His research aims at a sustainable agriculture in which the environment (such as climate change and biodiversity) and food production are in harmony. He conducts research on production, evolutionary ecology, and conservation of various crops and surrounding wild plants and animals, mainly in the field. He also conducts sustainable sumart agriculture with technologies such as drones and AI.

Wei Guo: He is an associate professor at the Graduate  School  of  Agricultural  and  Life  Sciences, The  University  of  Tokyo. His research interests include image sensing, machine learning, and plant phenotyping.

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.