2024 Volume 86 Issue 1 Pages 39-43
Novel object detection technology to discern the growth stages of sweet corn tassels was developed. Training of two models, the Coloring-model and the Flowering-model, was performed using YOLOv5. The models detected the growth stages of sweet corn tassels in three categories each: 1） Non-flowering, 2） Early stage of flowering and 3） Late stage of flowering for the Coloring-model, and 1） Non-flowering, 2） Partially flowering and 3） Fully flowering for the Flowering-model. Input images were taken by unmanned aerial vehicle （UAV） then resized in 640 or 1080 pixels before training. The results showed that the mean average precision of the Coloring-model and the Flowering-model are 0.71 and 0.61, respectively. In conclusion, these models were able to predict the expected harvest maturity.
