Disruptive technologies for smart farming in developing countries: Tomato leaf disease recognition systems based on machine learning

Abstract

Food security is a major concern in every developing country. Farmers face many problems while cultivating plants and they must take precautions at every stage of cultivation. Plants get diseases for various reasons like bacteria, insects, and fungus. Some diseases can be detected by examining the symptoms on the leaves. Early detection of diseases is a major concern and may require a thorough examination of the plants by an agricultural professional. This process is expensive and time taking. Machine learning (ML) algorithms help in image recognition and can be used to detect diseases on time without the need of an agricultural professional. In this project, the diseases in tomato leaves will be detected using image processing. The data from the images are extracted using different vectorization methods and classification algorithms like logistic regression (LR), support vector machine (SVM), and k-nearest neighbors (KNN). Vectors of size 32 × 32 and 64 × 64 are used for training with normalizer scaling and no scaling. Out of the different approaches that were explored, SVM with the radial basis function (RBF) kernel gives the highest accuracy of 85% with no scaling and 64 × 64 image dimension.