Cucumber Leaf Disease Classification using Local Tri-directional Patterns and Haralick Features

Plant's diseases cause major economic and production losses in the agricultural industry around the world. Farmers are mostly struggling in controlling and identification of plant diseases. Thus, timely detection of diseases will be beneficial for farmers, so they can avoid further losses. In this research, a machine learning-based technique is introduced for the classification of cucumber diseases i.e., Healthy, Anthracnose, Aphids and CYSDV. For conducting this research dataset is gathered from a cucumber farm with the single-lens reflex camera. The main aim is to design a model that is robust to noise in real-time problems. The immense noise in images made the task tough and interesting. Firstly, Images are converted to LAB color space, and Region of Interest (ROI) is extracted through K-mean clustering. Secondly, Haralick features, and Local Tri Directional pattern (LTriDP) are serially fused. Quadratic Support Vector Machine (Q-SVM) classifies disease with a success rate of 96.80%.