Rapid Nondestructive Detection of Welsh Onion, Onion, and Chinese Chives Seeds Based on Hyperspectral Imaging Technology
The appearance of <i>Allium</i> L. seeds is very similar, and it is difficult to achieve fast and accurate classification using traditional seed classification methods, which may cause damage to the seeds. Therefore, finding a quick and nondestructive classification method is very import...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-04-01
|
| Series: | Agriculture |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2077-0472/15/8/816 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The appearance of <i>Allium</i> L. seeds is very similar, and it is difficult to achieve fast and accurate classification using traditional seed classification methods, which may cause damage to the seeds. Therefore, finding a quick and nondestructive classification method is very important to solve the problem of seed confounding in actual production. In this study, hyperspectral imaging technology was combined with a variety of data preprocessing and classification models to achieve rapid and nondestructive classification of Welsh onion, onion, and Chinese chives seeds. In this paper, 1050 Welsh onion, onion, and Chinese chives seeds were used as materials, and their 400–1000 nm spectral images were collected for processing. Standard Normal Variable (SNV), Multivariate Scattering Correction (MSC), First-order Differential (FD), and Second-order Differential (SD) were used to denoise the spectral data. Then the dimensionality was reduced by Principal Component Analysis (PCA). Four classification models, Partial Least Squares Discriminant Analysis (PLS-DA), Support Vector Machine (SVM), Random Forest (RF), and k-Nearest Neighbor (KNN), were used to classify seeds quickly and accurately. The results show that the prediction accuracies of the Original-PLS-DA model, Original-Linear SVM model, and FD-Linear SVM model are the highest, reaching 98%, while the accuracy, recall rate, and F1 score all reach 96%. This study provides a new idea for rapid and nondestructive classification of <i>Allium</i> L. seeds in practical production. |
|---|---|
| ISSN: | 2077-0472 |