An Automatic-Segmentation- and Hyper-Parameter-Optimization-Based Artificial Rabbits Algorithm for Leaf Disease...

Area of Science:

• Agricultural Science
• Computer Science
• Artificial Intelligence

Background:

• Plant diseases pose significant threats to agriculture, causing substantial economic losses.
• Early detection and classification of plant diseases are crucial for minimizing spread and improving crop yields.
• Deep learning models require large datasets for accurate image classification, often facing overfitting challenges.

Purpose of the Study:

• To develop an advanced deep learning model for improved plant leaf disease classification.
• To address overfitting issues and enhance classification accuracy in plant disease detection.
• To introduce the Automatic Segmentation and Hyper Parameter Optimization Artificial Rabbits Algorithm (AS-HPOARA) for agricultural applications.

Main Methods:

• Utilized the Plant Village dataset for evaluating the AS-HPOARA approach.
• Applied Z-score normalization and three augmentation techniques (rotation, scaling, translation) to balance and preprocess images.
• Employed a modified UNet for image segmentation and HPO-based ARA for classification, including hyperparameter tuning.

Main Results:

• The AS-HPOARA model achieved a high classification accuracy of 99.7% for ten plant disease classes.
• Image augmentation techniques effectively reduced overfitting and improved classification accuracy.
• The proposed method demonstrated superior performance compared to existing models like CGAN-DenseNet121 and RAHC_GAN.

Conclusions:

• AS-HPOARA significantly enhances plant leaf disease classification accuracy.
• The developed algorithm effectively mitigates overfitting and improves model generalization.
• This approach holds potential for early disease detection, contributing to agricultural sustainability and economic stability.