Sweet potato appearance quality classification method and device based on transfer learning

The invention discloses a sweet potato appearance quality classification method and device based on transfer learning. A sweet potato image data set is constructed through the RGB images of classified sweet potatoes, and a sweet potato appearance quality classification network is designed through th...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: LI JINGBING, LI WEIMING, GONG CHENG, SHENG MINGSHUAI, ZHANG WENXING
Format: Patent
Sprache:chi ; eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The invention discloses a sweet potato appearance quality classification method and device based on transfer learning. A sweet potato image data set is constructed through the RGB images of classified sweet potatoes, and a sweet potato appearance quality classification network is designed through the combination of a feature extraction part of a trained ResNet18 deep convolutional neural network and a classifier. The sweet potato image data set is used for training the sweet potato appearance quality classification network, so that when the sweet potato appearance quality classification network is used for detecting and classifying the sweet potato appearance quality, the detection efficiency is improved, more accurate detection precision is achieved, the appearance quality classification network is used for detecting and classifying the sweet potatoes, labor force is released and labor cost is saved, and damage of human factors to the sweet potatoes is reduced. The invention can solve the technical problems