AI-driven innovations for onion seed producers

Onion seed production is a sophisticated process that faces numerous challenges, including climate variability, pests, diseases, and the necessity for sustainable agricultural practices. A critical aspect of this process is accurately counting onion umbels, as this allows growers to estimate the potential yield of their crops and devise more effective crop management strategies. 

Onion umbels counting - use case background

An onion producer managing a 7-hectare field in Zaragoza, Aragon, sought an innovative solution to enhance the accuracy of yield evaluations. The producer was facing difficulties in accurately identifying umbels and distinguishing them from other objects in the field, such as rocks. 

Luis Valdez, Proofminder’s representative in Spain, led the project, working closely with the producer to understand their specific challenges and needs. His expertise in agricultural technology and deep understanding of local farming practices were instrumental in tailoring the solution. 

The process

To address these challenges, images of the field were collected using a DJI Phantom 4 RTK drone, providing high-resolution data at 0.5 GSD. Proofminder then utilized this data to train an AI model capable of: 

• Counting onion umbels. 

• Classifying umbels by size (small, medium, and large). 

• Delivering a hyper-accurate report to the producer via a web application.

Figure 1 – Proofminder’s AI model for onion seed growers

Figure 2 – Onion umbels counting and classified by size (small, medium, and large)

Challenges in the innovation process

During the development phase, it became apparent that the AI model needed to differentiate between onion umbels and white rocks in the field. To solve this, the model was fine-tuned to recognize rocks separately, achieving an impressive accuracy rate of over 95%. This additional feature not only enhanced the accuracy of umbel counts but also contributed to safer cultivation practices by helping the grower manage the presence of rocks in their field. 

The results

• Accurate Yield Assessment: The technology enables a more precise evaluation of crop yields. 

• Cross-referencing Information: The model’s accuracy can be validated through sample counts or during harvest. 

• Evaluation of Productive Potential: The field can be zoned more effectively, allowing for a more accurate assessment of its productive potential. 

• Additional Insights: The model provides additional data about the field and plants, aiding in better decision-making, such as: 

• Onion umbels were categorized by size (small, medium, large). 

• Results allow the grower to compare the yield with that of previous crop years. 

• If the field is rented, it allows the grower to make an educated decision on whether to discard or ratify the work with the same field. 

• Digitized Information: All information is presented in a digitized format, easily accessible via a web application. 

A Look Ahead

This AI model is versatile and can also be applied to other crops such as garlic, leek, watermelon, melon, pumpkin, carrot, parsley, fennel, coriander, and dill. 

As the agricultural landscape continues to evolve, the integration of AI technologies is no longer a luxury but a necessity for competitive and sustainable farming. By adopting innovative solutions like AI-driven umbel counting models, onion seed producers can optimize their yields, reduce risks, and make data-driven decisions that enhance overall productivity. 

If you’re a grower looking to elevate your farming practices with cutting-edge technology, don’t hesitate to explore how Proofminder’s AI models can be tailored to your specific needs.