An Advanced Machine Learning Framework for Apple Tree Leaf Disease Identification: Integrating MS-LBP, HOG, and MRMR Feature Selection with XGBoost Classification

Abstract

: The agricultural industry faces major economic hurdles from apple tree leaf diseases when these diseases remain undetected or unmanaged in their timely manner. The research develops an effective method which employs image processing and machine learning techniques to detect and classify diseases in apple tree leaves. A combination of MSLBP features with HOG features helps extract extensive texture features that the method combines into one unified vector. Minimum Redundancy Maximum Relevance (MRMR) selects the most disease-diagnostic features from the vector through a thorough evaluation process. XGBoost classifier along with robust ensemble features achieves classification of selected features which exhibits high performance values. The novel approach contains feature selection and integration that leads to 98.03% classification accuracy compared to the 97.56% accuracy of the baseline method. The system provides improved accuracy alongside solid performance across different environmental conditions which makes it an optimal solution for real-time disease management in apple tree leaf detection for precision agriculture.