Disease diagnostic method based on cascade backbone network for apple leaf disease classification.

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Fruit tree diseases pose significant agricultural challenges in China.
• Mobile devices are increasingly used for agricultural pest and disease identification.
• Accurate and efficient identification of apple leaf diseases is crucial for crop management.

Purpose of the Study:

• To develop an efficient and accurate method for identifying apple leaf diseases using mobile devices.
• To introduce a novel cascade backbone network (CBNet) incorporating MobileViT for enhanced feature extraction.
• To improve the detection of apple leaf diseases in field conditions.

Main Methods:

• Proposed a novel cascade backbone network (CBNet) for apple leaf disease identification.
• Integrated MobileViT-based convolutional blocks for superior feature extraction.
• Developed a feature refinement module and employed a pyramidal cascaded multiplication operation for feature fusion.

Main Results:

• The proposed CBNet achieved an accuracy of 96.76% and an F1-score of 96.71% on field datasets.
• MobileViT-based blocks demonstrated enhanced capability in mining image feature information.
• The feature refinement module and cascaded fusion improved detection performance.

Conclusions:

• The developed CBNet method offers a promising solution for automated apple leaf disease identification.
• This research is the first to apply Transformer technology to apple leaf disease identification, yielding significant results.
• The method demonstrates high potential for practical application in agricultural settings using mobile devices.