Machine Learning-Based for Automatic Detection of Corn-Plant Diseases Using Image Processing

Abstract

Corn is one of the major crops in Sudan. Disease outbreaks can significantly reduce maize production, causing huge damage. Conventionally, disease diagnosis is made through visual inspection of the damage in fields or through laboratory tests conducted by experts on the affected plant parts of the crop. This process typically requires highly skilled personnel, and it can be time-consuming to complete the necessary tasks. Machine learning methods can be implemented to rapidly and accurately detect disease and reduce the risk of crop failure due to disease outbreaks. This study aimed to use traditional machine learning techniques to detect maize diseases using image processing techniques. A total of 600 images were obtained from the open-source Plant Village dataset for experimentation. In this study, image segmentation was done using K-means clustering, and a total of 4 GLCM texture features and two statistical features were extracted from the images. In this study, four traditional machine learning algorithms were applied to detect diseased maize leaves (common rust and gray leaf spot) and healthy maize leaves. The results showed that all the algorithms performed well in identifying the diseased and healthy leaves, with accuracy rates ranging from 90% to 92.7%. The highest accuracy scores were obtained with support vector machine and artificial neural networks, respectively.