| Autor: |
Yang, Jiwen, Wang, Guohui |
| Předmět: |
|
| Zdroj: |
Journal of Food Measurement & Characterization; Dec2023, Vol. 17 Issue 6, p5794-5805, 12p |
| Abstrakt: |
Cherries are spring fruits enriched with nutrients and sweetness, which are widely popular among consumers all over the world. Cherry appearance characteristic is an important indicator for consumers' purchases and robot picking, which is closely related with classifying cherry maturity and monitoring cherry diseases. It is therefore very important to accurately identify the overall appearance quality of cherries. Fruit appearance quality detection has been continuously developed by advanced deep learning techniques. In this paper, an innovative cherry appearance detection method is developed for rapid and accurate classification of different cherry maturity stages and disease status. The different convolutional neural networks are used as a feature extractor to extract the deep features and then the features are imported into nine classifiers to detect the appearance quality of cherries. The fine-tuned method performs quite well with superior accuracy of cherry classification at less computation overhead for training. The experimental results show that: (1) Among the 54 combined models, the accuracy of twenty-two combined models is higher than transfer learning. (2) The training time of all the combined models is tremendously shorter than transfer learning. (3) The best model, Mobilenet_v2 plus support vector machine, have a recognition accuracy of 98.3% and the training time is 73.478 s. The experimental results show that the method of extracting features into classifiers has a good effect on recognizing cherry appearance quality. The successful application of this method provides a new solution for cherry appearance ripeness and disease identification. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
Full text from SpringerLink