Segmented hue saturation value histogram model for leaf disease diagnosis and classification using computer vision

Various plants leaf is used in daily life for cooking, medicine etc. As is well known, undesirable organisms, such as diseases, cause a large portion of the yield of agricultural crops to be lost. The identification of plant diseases is constantly improving, and new techniques and technologies are being developed and implemented to ensure the precision and speed of diagnosis. They use advanced tools and technologies to diagnose plant diseases in a timely, efficient, and highly accurate manner. This facilitates accurate and timely decision-making regarding the control of crop diseases in agriculture. From the last few years, the computer vision methods have been used in identification and classification of healthy and diseased leaves. The proposed framework first extracts the features of leaves using segmented HSV histogram analysis then the features taken out are fed to random forest, logistic regression and support vector machine for identification as healthy or having disease in them. The framework is tested on the plant village dataset (Grape’s leaf) to identify three diseases a) black rot b) black measles c) leaf blight. With fine tuning of parameters, the result showed that SVM achieved 95.41% recognition accuracy, logistic regression achieved 96.11% and random forest achieved 94.17% for the model. The proposed design is compared with three CNN architecture [ResNet50, EfficientNetB0, InceptionV3] without fine tuning. It is found that, ML models are computationally efficient and maintain high accuracy, making them suitable for low-resource applications such as mobile-based or edge AI systems.