Title: Investigation of different feature extraction methods for intelligent diagnosis and classification of plant leaf diseases

Abstract

Plant diseases are serious causes in reducing quality and quantity of productions. Visual evaluation of plants by human observers is time consuming, costly and prone to error. Disease assessment and plant maintenance require new and innovative methods to meet the challenges in the field of agricultural production. In this regard, sensors and imaging techniques have shown great potential in creating new approaches to plant pathology interactions and the diagnoses of plant diseases. Advances in agricultural technology have created opportunities for the diagnosis and non-destructive classification of plant diseases. There are many advances in computer vision that help identify and classify plant diseases automatically. A classifier diagnoses plant as healthy and unhealthy with the given features (color, texture, and shape) as input to automatic diagnosis. Accuracy is the main parameter that every researcher uses to calculate the performance of the model. The accuracy of the classifier depends primarily on the features that are extracted; therefore, feature extraction plays a vital role in identifying a disease. Proper selection of the correct features leads to high diagnostic accuracy. Feature engineering in Machine Learning and Deep Learning are the two main types of feature extraction methods which will be explored in this research.