We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Integration of Optical Property Mapping and Machine Learning for Real-Time Classification of Early Bruises of Apples.
- Authors
Hu, Dong; Qiu, Dekai; Yu, Shengqi; Jia, Tianze; Zhou, Tongtong; Yan, Xiaojie
- Abstract
Real-time detection and classification of bruised apples are critical to appropriate postharvest handling. However, the early bruises are hard to identify by vision inspection due to no or few visual symptoms. In this study, the feasibility of optical property mappings was investigated for the real-time classification of early apple bruises (occurring beneath the peel for 0 h), coupled with machine learning methods. Spatial-frequency domain images of 360 apples at multiple spatial frequencies with three phases were first acquired, followed by image demodulation and inverse estimation for generating optical property mappings, through which the bruised apple tissues were visualized. Then, support vector machine (SVM) and convolutional neural network (CNN) were used to distinguish two-category (intact and bruised) and three-category apples (intact, mildly bruised, and severely bruised). The best classification accuracy of the prediction set for the two-category using SVM was up to 98.33%. CNN for the classification of bruises outperformed SVM, with the accuracies of 99.16% and 91.50% for the two-category and the three-category, respectively. However, the time consumed for CNN classification was three orders of magnitude more than the SVM. Dynamic changes in the optical property mappings of the apples after bruising within 24 h demonstrated that classification accuracy would likely increase with the bruising time. Comparison between our results and other literature further validated the feasibility of the proposed method for real-time classification of apple early bruise.
- Subjects
CONVOLUTIONAL neural networks; SUPPORT vector machines; MACHINE learning; OPTICAL properties; DEMODULATION
- Publication
Food & Bioprocess Technology, 2024, Vol 17, Issue 9, p2745
- ISSN
1935-5130
- Publication type
Article
- DOI
10.1007/s11947-023-03260-5