Authenticity Recognition of Rice Varieties Based on Polarization Imaging Technology
To develop a non-destructive and efficient image recognition technology to verify rice variety authenticity, this study centered on three closely related japonica rice varieties and three indica rice varieties. Visible light images and four distinct types of polarized images (polarization intensity...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | zho |
| Published: |
The editorial department of Science and Technology of Food Industry
2025-04-01
|
| Series: | Shipin gongye ke-ji |
| Subjects: | |
| Online Access: | http://www.spgykj.com/cn/article/doi/10.13386/j.issn1002-0306.2024040326 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850095541089730560 |
|---|---|
| author | Jie FU Yue WU |
| author_facet | Jie FU Yue WU |
| author_sort | Jie FU |
| collection | DOAJ |
| description | To develop a non-destructive and efficient image recognition technology to verify rice variety authenticity, this study centered on three closely related japonica rice varieties and three indica rice varieties. Visible light images and four distinct types of polarized images (polarization intensity I0° image, polarization Stokes vector S0 image, polarization angle image, and polarization degree images) were gathered for each rice type. Utilizing six convolutional neural network algorithms (AlexNet, VGG16, GoogLeNet, ResNet34, DenseNet, and ConvNeXt V2), models were established to identify the authenticity of japonica and indica rice varieties based on various image types and algorithms. When comparing the accuracy of these models on validation sets, it was observed that for identifying the authenticity of japonica rice varieties, the ResNet network based on visible light images achieved the highest accuracy at 100%, while the VGG16 network based on polarization degree images attained 98.5%. In the case of identifying the authenticity of indica rice varieties, the VGG16 network using polarization Stokes vector S0 images recorded the highest accuracy of 99.5%, while both the VGG16 and ResNet networks using polarization degree images achieved an accuracy of 99.3%. This study highlights the practical feasibility of employing polarization imaging technology for authenticating rice varieties and offers valuable reference data for selecting appropriate image types and algorithms for japonica and indica rice variety identification. |
| format | Article |
| id | doaj-art-21ef45eaca9b4cac802481dfaa07b155 |
| institution | DOAJ |
| issn | 1002-0306 |
| language | zho |
| publishDate | 2025-04-01 |
| publisher | The editorial department of Science and Technology of Food Industry |
| record_format | Article |
| series | Shipin gongye ke-ji |
| spelling | doaj-art-21ef45eaca9b4cac802481dfaa07b1552025-08-20T02:41:25ZzhoThe editorial department of Science and Technology of Food IndustryShipin gongye ke-ji1002-03062025-04-0146723524710.13386/j.issn1002-0306.20240403262024040326-7Authenticity Recognition of Rice Varieties Based on Polarization Imaging TechnologyJie FU0Yue WU1National Engineering Laboratory for Deep Processing of Rice and Byproduct, College of Food Science and Engineering, Central South University of Forestry and Technology, Changsha 410004, ChinaNational Engineering Laboratory for Deep Processing of Rice and Byproduct, College of Food Science and Engineering, Central South University of Forestry and Technology, Changsha 410004, ChinaTo develop a non-destructive and efficient image recognition technology to verify rice variety authenticity, this study centered on three closely related japonica rice varieties and three indica rice varieties. Visible light images and four distinct types of polarized images (polarization intensity I0° image, polarization Stokes vector S0 image, polarization angle image, and polarization degree images) were gathered for each rice type. Utilizing six convolutional neural network algorithms (AlexNet, VGG16, GoogLeNet, ResNet34, DenseNet, and ConvNeXt V2), models were established to identify the authenticity of japonica and indica rice varieties based on various image types and algorithms. When comparing the accuracy of these models on validation sets, it was observed that for identifying the authenticity of japonica rice varieties, the ResNet network based on visible light images achieved the highest accuracy at 100%, while the VGG16 network based on polarization degree images attained 98.5%. In the case of identifying the authenticity of indica rice varieties, the VGG16 network using polarization Stokes vector S0 images recorded the highest accuracy of 99.5%, while both the VGG16 and ResNet networks using polarization degree images achieved an accuracy of 99.3%. This study highlights the practical feasibility of employing polarization imaging technology for authenticating rice varieties and offers valuable reference data for selecting appropriate image types and algorithms for japonica and indica rice variety identification.http://www.spgykj.com/cn/article/doi/10.13386/j.issn1002-0306.2024040326ricevarietiesauthenticity identificationpolarization imaging techniquesdeep learning |
| spellingShingle | Jie FU Yue WU Authenticity Recognition of Rice Varieties Based on Polarization Imaging Technology Shipin gongye ke-ji rice varieties authenticity identification polarization imaging techniques deep learning |
| title | Authenticity Recognition of Rice Varieties Based on Polarization Imaging Technology |
| title_full | Authenticity Recognition of Rice Varieties Based on Polarization Imaging Technology |
| title_fullStr | Authenticity Recognition of Rice Varieties Based on Polarization Imaging Technology |
| title_full_unstemmed | Authenticity Recognition of Rice Varieties Based on Polarization Imaging Technology |
| title_short | Authenticity Recognition of Rice Varieties Based on Polarization Imaging Technology |
| title_sort | authenticity recognition of rice varieties based on polarization imaging technology |
| topic | rice varieties authenticity identification polarization imaging techniques deep learning |
| url | http://www.spgykj.com/cn/article/doi/10.13386/j.issn1002-0306.2024040326 |
| work_keys_str_mv | AT jiefu authenticityrecognitionofricevarietiesbasedonpolarizationimagingtechnology AT yuewu authenticityrecognitionofricevarietiesbasedonpolarizationimagingtechnology |