White strawberries are high-value agricultural products whose ripeness identification poses a notable challenge for robotic harvesting. This study first evaluated YOLOv5 and YOLOv12 for vision-based ripeness detection, using conventional strawberries as a control group. Results showed that both algorithms yielded substantially lower mAP scores on white strawberries than on conventional ones, with a maximum of only 68.20%, confirming the inadequacy of current image-based approaches. To overcome this limitation, industrial-grade infrared cameras were employed to acquire infrared images and gray-scale intensity curves from six groups of white and conventional strawberries under laboratory and field conditions. Validation through peak-contrast gray-scale analysis, least-squares Gaussian fitting, linear discriminant analysis, and principal component analysis demonstrated that the infrared gray-scale curve method reliably differentiated ripeness stages for both strawberry types, outperforming conventional visual recognition. These findings provide a practical basis for automated harvesting and ripeness assessment systems.
Jingjing et al. (Mon,) studied this question.