|
Research Article: Fish freshness detection from eye images using feature-based fusion and support vector machine (SVM) classification
|
A. Zare Kordkheili1   , M. Nasrolahzadeh2 , S. Asadi Amiri1 , Z. Mohammadpoory3 , A. Movahedinia *4 |
1- Department of Computer Engineering, University of Mazandaran, Babolsar, Iran
2- Department of Electrical Engineering, University of Torbat Heydarieh, Torbat Heydarieh, Iran
3- Department of Biomedical Engineering, Shahrood University of Technology, Shahrood, Iran
4- Department of Marine Biology, Faculty of Marine and Environmental Sciences, University of Mazandaran, Babolsar, Iran , amovahedinia@umz.ac.ir |
|
|
Abstract: (24 Views) |
| Fish is a staple food for people, and ensuring its freshness is crucial for the industry. Various parameters, including fish eye characteristics, gill features, and fish fins, are commonly used to distinguish fish quality. In this study, we propose a novel method to assess fish freshness using fish eye images. Initially, data augmentation is employed to increase the effective size of the training dataset, enhancing robustness to variations, balancing class distributions, and reducing overfitting. In the proposed method, we utilized three convolutional neural networks: Inception-v3, VGG16, and MobileNetV3, to detect fish spoilage. We made slight structural modifications to each of these networks to enhance their performance in detecting fish freshness. In addition, we extracted feature vectors from the global average pooling layer of each network. We then used a Support Vector Machine (SVM) to classify the freshness of the fish. This study utilized the Freshness of Fish Eyes (FFE) dataset, which includes 8 species of fish at 3 levels of freshness. The proposed method, using Inception-v3 and the SVM classifier, achieved an accuracy of 81.21%, which is 4% better than the existing method on this dataset. This method provides a significant advancement in fish freshness assessment, offering a more accurate and reliable means of determining fish quality. This can greatly benefit the food industry by ensuring higher standards of freshness, reducing waste, and improving consumer satisfaction. The demonstrated improvement in accuracy highlights the potential of this method to set new benchmarks in fish quality assessment. |
|
| Keywords: Fish quality detection, Inception-v3, VGG16, MobileNet |
|
|
Full-Text [PDF 1511 kb]
(14 Downloads)
|
Type of Study: Orginal research papers |
Subject:
fish disease
ePublished: 2026/03/17
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
