A Machine Vision Approach for Nitrate Sufficiency Determination in Aquaponic Lettuce via Leaf Color Recognition

Aquaponic systems are popular as sustainable food production methods, however it also experiences challenges in terms of monitoring the nutrient levels, specifically nitrate sufficiency. In this study a new method is proposed to monitor the nitrate sufficiency in aquaponic systems using leaf color recognition. To do this, a state-of-the-art deep learning algorithm is utilized, the YOLOv8 for its extensive object detection capabilities. This algorithm is used to generate a custom model for leaf color recognition. The model is trained to recognize whether a lettuce leaf is nitrate deficient or sufficient based on its color. The model attained an amazing mean average precision (mAP) of 99.4% after extensive training and evaluation. The high mAP score implies that the model is highly accurate and reliable in differentiating between leaves with deficient and sufficient nitrate levels. With this, farmers and researchers may efficiently monitor and manage nitrate levels through integrating this leaf color recognition technology into aquaponic setups.