VGG-MFO-orange for sweetness prediction of Linhai mandarin oranges DOI Creative Commons

Chun Fang,

Runhong Shen,

Meng-Di Yuan

и другие.

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Апрель 6, 2025

Mandarin orange is a popular fruit in China and known worldwide for its unique flavor nutritional benefits. As consumer demand quality increases, the fine assessment grading of sweetness-especially through non-destructive testing techniques-are becoming increasingly important agriculture commerce. In this paper, new Attention Orange (AO) attention mechanism Multiscale Feature Optimization (MFO) feature extraction module are designed combined with VGG13 convolutional neural network (CNN), innovatively proposed VGG-MFO-Orange CNN model accurately classifying mandarin oranges different sweetness. First, sample Linhai was collected, sweetness triple classification dataset 5022 images formed, utilizing image acquisition sugar detection. The then trained against six influential classical models: DenseNet121, MobileNet_v2, ResNet50, ShuffleNet, VGG13, VGG13_bn. experimental results showed that our achieved an accuracy 86.8% on validation set, which significantly better than other models. It also demonstrated excellent generalization ability effectiveness predicting oranges. Therefore, can provide efficient means agricultural production, contribute to modernization, enhance competitiveness products market.

Язык: Английский

VGG-MFO-orange for sweetness prediction of Linhai mandarin oranges DOI Creative Commons

Chun Fang,

Runhong Shen,

Meng-Di Yuan

и другие.

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Апрель 6, 2025

Mandarin orange is a popular fruit in China and known worldwide for its unique flavor nutritional benefits. As consumer demand quality increases, the fine assessment grading of sweetness-especially through non-destructive testing techniques-are becoming increasingly important agriculture commerce. In this paper, new Attention Orange (AO) attention mechanism Multiscale Feature Optimization (MFO) feature extraction module are designed combined with VGG13 convolutional neural network (CNN), innovatively proposed VGG-MFO-Orange CNN model accurately classifying mandarin oranges different sweetness. First, sample Linhai was collected, sweetness triple classification dataset 5022 images formed, utilizing image acquisition sugar detection. The then trained against six influential classical models: DenseNet121, MobileNet_v2, ResNet50, ShuffleNet, VGG13, VGG13_bn. experimental results showed that our achieved an accuracy 86.8% on validation set, which significantly better than other models. It also demonstrated excellent generalization ability effectiveness predicting oranges. Therefore, can provide efficient means agricultural production, contribute to modernization, enhance competitiveness products market.

Язык: Английский

Процитировано

0