Integration of Deep Learning and Machine Learning Techniques for Advancing the Detection of Plant Diseases

Abstract

This research proposed a new hybrid system for the detection of diseases in Pepper, Tomatoes, and Potatoes vegetable crops. Vegetable plant disease poses a significant threat to both the quality and quantity of crop yields, leading to enormous economic losses in the agricultural industry. Early detection and prompt measures are essential for the effective control of plant diseases. Extreme Gradient Boosting (XGBoost) and Convolutional Neural Network (CNN) algorithm were employed to detect diseases precisely. CNN derived intricate patterns from images, whereas the XGBoost algorithm is employed for precise categorization. The model is built for Pepper, Tomatoes, and Potatoes diseases and also for plant type classification using the dataset which was captured by mobile phone camera from Kotri and Unerpur, Sindh, Pakistan with accuracy of 93%, 92%, 99%, and 98%, respectively. Detection of new diseases which were not detected earlier is another primary aim of this research. For the detection of the unknown kind of diseases in plants, the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm was utilized for clustering the similar captured images and for disease classification; then CNN XGBoost was applied on the newly developed dataset. This study employed a new automated hybrid framework that combines the strengths of CNN and XGBoost to identify familiar and unfamiliar diseases in vegetable crops.