International Conference on Advanced Network Technologies and Applications (APAN)
Quantum Support Vector Machines for Coffee Leaf Disease Classification Using PCA-Reduced Texture Features
Sourav Samanta, Ramkumar Natarajan
Abstract
The application of machine learning in the era of digital agriculture has rapidly transformed the Indian agricultural landscape. Simultaneously, quantum technology has begun to play a crucial role in modern computational paradigms, leading to the research of Quantum Machine Learning (QML). In recent years, QML has been applied to solve complex challenges in modern sectors, including smart agriculture. The Quantum Support Vector Machine (QSVM) is a quantum-enhanced variant of the classical Support Vector Machine (SVM). This research investigates the performance of QSVM in binary classification of coffee diseases, specifically coffee leaf rust and phoma. Six Gray-Level Co-occurrence Matrix (GLCM) features: contrast, correlation, dissimilarity, energy, entropy, and homogeneity were extracted. PCA was applied to the combined 18 features, and varying numbers of components were extracted. Then classification was performed with both SVM and QSVM on the reduced feature vector. In the experiment, three SVM kernels, three QSVM feature maps, and five classification evaluation metrics were used to conduct a comprehensive study. The experimental results show that the SVM with a polynomial kernel achieved 89.83% accuracy, while the QSVM with a full feature map achieved 95.83% accuracy in classifying healthy leaves from those affected by coffee leaf rust. Additionally, the SVM with a polynomial kernel achieved an accuracy of 83.17% in classifying healthy leaves from those affected by phoma, whereas the QSVM achieved 96.17%. Finally, the research discussed the performance of the QSVM under different evaluation parameters and its limitations.