Automated In Situ Seed Variety Identification via Deep Learning: A Case Study in Chickpea

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Accurate seed variety recognition is crucial for preventing yield loss and ensuring synchronized crop production.
• Traditional seed identification methods are subjective, error-prone, and require expert knowledge and accredited materials.

Purpose of the Study:

• To develop an automated system for chickpea seed variety identification using a convolutional neural network (CNN).
• To assess the model's robustness across variable image acquisition conditions and devices.

Main Methods:

• A modified VGG16 convolutional neural network (CNN) architecture was employed for image analysis.
• Images of chickpea seeds were captured using low-cost devices under varying lighting and imaging settings.
• A five-fold cross-validation was utilized to evaluate the model's predictive performance and uncertainty.

Main Results:

• The developed deep learning model accurately identified diverse chickpea seed varieties with an average classification accuracy exceeding 94%.
• The vision-based model demonstrated high robustness, performing consistently regardless of the image acquisition device, lighting, or imaging parameters.
• The model successfully distinguished intricate visual features among different chickpea varieties.

Conclusions:

• The proposed CNN framework provides a reliable and automated solution for chickpea seed variety identification.
• The model's robustness and accuracy pave the way for in situ applications using mobile devices.
• This technology offers significant potential for deployment in the seed industry, enhancing automated identification practices.