Vegetable and fruit freshness detection based on deep features and principal component analysis.

Area of Science:

• Agricultural Science
• Computer Science
• Food Science

Background:

• Current manual methods for vegetable and fruit freshness detection are subjective and inaccurate, hindering large-scale, efficient assessment.
• Artificially extracted features in existing automated methods show poor adaptability, limiting freshness detection efficiency.
• Objective, accurate, and efficient freshness detection is crucial for consumer health, industry quality, and market competitiveness.

Purpose of the Study:

• To develop a novel method for objective, accurate, and efficient vegetable and fruit freshness detection.
• To leverage deep features from pre-trained deep learning models for enhanced freshness assessment.
• To improve the adaptability and efficiency of automated freshness detection systems.

Main Methods:

• Deep features were extracted from resized vegetable and fruit images using pre-trained deep learning models (GoogLeNet, DenseNet-201, ResNeXt-101).
• Extracted deep features were fused and reduced in dimensionality using Principal Component Analysis (PCA).
• Freshness detection was performed using three distinct machine learning classifiers on the reduced feature set.

Main Results:

• The combined deep features from GoogLeNet, DenseNet-201, and ResNeXt-101, after PCA dimensionality reduction, achieved the highest accuracy.
• The proposed method reached an accuracy rate of 96.98% for vegetable and fruit freshness detection.
• The results demonstrate superior performance compared to traditional feature extraction methods.

Conclusions:

• The proposed deep learning-based method significantly improves the objectivity, accuracy, and efficiency of vegetable and fruit freshness detection.
• Combining features from multiple pre-trained deep learning architectures enhances detection performance.
• This approach shows strong potential for practical application in the agricultural and food industries.