Dual-Stage Plant Disease Detection Using Image Processing and AI: Species Classification with SVM and Disease Identification with CNN

The study presents a robust approach for plant disease detection leveraging image processing and artificial intelligence. Initially, the system classifies the plant species using a comprehensive plant disease dataset. An input image undergoes pre-processing to remove noise and restore clarity, followed by feature extraction using Histogram of Oriented Gradients (HOG) features. This data feeds into a Support Vector Machine (SVM) classifier, identifying the plant as one of several types: Apple, Corn, Grape, Pepper Bell, Potato, or Tomato. Subsequently, the identified plant image is subjected to disease classification through a six-layer Convolutional Neural Network (CNN). Each plant type has a set of specific diseases: for Apple, the diseases are Black Rot, Scab, and Cedar-Rust; for Corn, they are Common Rust, Northern Blight, and Cercospora; for Grapes, the diseases include Black Rot, Measles, and Leaf Blight; and for Potato, the disease is Late Blight. The CNN model diagnoses the disease and recommends appropriate fertilizers for treatment. This dual-stage classification system enhances the precision of disease identification and provides practical agricultural advice, significantly aiding in the effective management and treatment of plant diseases.

Keywords: Plant Disease Dataset, Pre-Processing, Convolutional Neural Networks, Deep learning, machine learning, Classification, Accuracy.