Dynamic quality governance of avocado ripeness: cyber-physical system integrating multi-sensor networks and adaptive control

Perception and evaluation of ripeness in post-ripening fruit is complex, relying mainly on external color, but color alone is difficult to assess fully and accurately. Internal qualities such as soluble solids, hardness, pH, and respiration rate are also changing, and the lack of a clear stage division makes perception and evaluation difficult. Therefore, this study proposes a method of ripening regulation and perception based on multiple environmental parameters designed based on fruit ripening characterization indexes and warehouse environmental parameters. YOLOv5 (You Only Look Once Version 5) was used to classify the ripeness of fruit appearance with an accuracy of 98%. Then, through the establishment of LSTM (Long Short-Term Memory Network) level 1 and XGBoost (Extreme Gradient Boosting) level 2 ripeness regulation and perception evaluation models, the correlation between multiple parameters and ripeness indicators is explored and the ripeness is predicted, to realize the ripeness regulation and perception assessment based on multiple parameters. The study showed that the classification accuracy of the multi-environmental parameter-based ripeness perception and regulation model reached 93%, and the regression prediction R2 reached 0.94, which was higher compared with the traditional artificial perception of ripeness and other models. This study provides a ripening control and perception method based on multi-environmental parameters for the agricultural field, which is expected to build a post-harvest ripeness assessment and perception system for avocados in agricultural production and can be promoted for other fruit.

Text 123668.pdf - Published Version Restricted to Repository staff only Download (7MB) Official URL or Download Paper: https://link.springer.com/article/10.1007/s11694-0...

Actions (login required)

View Item