Identification of Alfalfa Leaf Diseases Using Image Recognition Technology

Area of Science:

• Agricultural science
• Plant pathology
• Computer vision
• Machine learning

Background:

• Four common alfalfa leaf diseases, including common leaf spot, rust, Leptosphaerulina leaf spot, and Cercospora leaf spot, significantly impact crop quality and yield.
• Timely and accurate disease diagnosis is crucial for effective management strategies, ensuring the health of alfalfa crops and the sustainability of the alfalfa industry.

Purpose of the Study:

• To investigate the identification and diagnosis of four common alfalfa leaf diseases using pattern recognition algorithms and image-processing technology.
• To develop and compare machine learning models for accurate alfalfa leaf disease detection based on extracted image features.

Main Methods:

• Digital images of alfalfa leaves were processed, and lesions were segmented using a combination of K-median clustering and linear discriminant analysis.
• Texture, color, and shape features were extracted from segmented lesion images (129 features total).
• Feature selection was performed using ReliefF, 1R, and correlation-based methods, followed by model building with Support Vector Machine (SVM), random forest, and K-nearest neighbor algorithms.

Main Results:

• The optimal model was an SVM classifier using the top 45 features selected by ReliefF, achieving 97.64% accuracy on the training set and 94.74% on the testing set.
• Semi-supervised models, utilizing the same 45 features, achieved approximately 80% accuracy for both training and testing sets.
• The study demonstrated high accuracy in image recognition for the four investigated alfalfa leaf diseases.

Conclusions:

• Image recognition technology, combined with advanced machine learning algorithms, provides a feasible and accurate solution for diagnosing common alfalfa leaf diseases.
• The developed methodology offers a valuable tool for disease management, quality control, and supporting the alfalfa industry.