Advancing Pear Disease Recognition: A Hybrid CNN-Random Forest Approach

The pear is a key fruit in the horticultural sector, renowned for its adaptability and nutritional value. In recent years, the effects of climate change have increased dramatically. It is influencing all stages of plant production and compelling farmers to adopt new technologies based on data analytics to modify and adapt their crop management practices. Every year, plant diseases are proven to cause significant economic losses. This study introduces a novel method for automating the recognition of pear disease using a synergistic combination of Convolutional Neural Networks (CNN) and Random Forest. A meticulously collected and annotated dataset containing 3,150 high-resolution images of pear leaves in various disease states was compiled. The proposed system's exceptional accuracy of 77.2% signifies a significant advance in agricultural technology. The CNN module is optimised for robust feature extraction, which enables the model to recognise intricate patterns and textures that are characteristic of various pear diseases. Utilising the collective intelligence of a collection of decision trees, a Random Forest classifier is used to refine classification. This study demonstrates the potential of deep learning-based models in the detection and handling of pear disease, which can result in substantial enhancements to pear production and food security.