The proposed system aims to automate soybean classification stages for cooperatives. Currently, many locations perform this process manually, which can lead to errors due to human factors. Therefore, SOYA: Automated Soybean Grain Classifier (suggested cool name: "GrainScan AI" or "SoyaSort Pro") seeks to automate the classification process in compliance with the Ministry of Agriculture's regulations.

How It Works

1. Sample Input – The classifier must insert a soybean sample of at least 125g into the machine.
2. Pre-Screening – A sieve will filter out large impurities (e.g., leaves and twigs).
3. Fine Sieving – The remaining material will pass through a 3mm sieve to separate grains from smaller particles.
4. Weight Measurement –
   • The material that passes through the sieve will be weighed.
   • The retained grains will be dropped into a container and also weighed.
5. AI-Based Classification – A camera will take images of the grains in the container to classify them based on the following characteristics:
   • Good grains
   • Broken grains
   • Crushed grains
   • Burnt grains
   • Sprouted grains
   • Moldy grains
6. Report Generation – The trained AI model will generate a classification report following the Ministry of Agriculture’s standards.
7. Moisture estimation - After the images have been taken, the grains will be deposited in an output compartment that will estimate the moisture content of the grains.

[References]

https://www.sistemafaep.org.br/wp-content/uploads/2024/01/PR.0230-Classificacao-Soja_web.pdf