The fusion and comparison of image processing algorithms for prediction of leaf disease

Many bacteria, fungi, and viruses damage a plethora of plants in the United States. More than sixty percent of the total U.S. crop losses are due to pathogens and fungi, adding up to an estimated expense of 137 billion dollars a year. Leaf diseases are a significant threat to humans as consumers and breeders of the plants. Plant quality and productivity are affected by the different types of diseases. Many are infectious and can cause a total loss of crop yield Timely and accurate detection of plant diseases is essential for crop quality and yield. Early detection can reduce plant disease; farmers rely on experts to identify most of the diseases. For a small farm measuring a few acres, the traditional method of a disease detection may be possible, but this method is not feasible and is costly for a large field. In many rural areas, farmers often lack precise knowledge or accessibility to an expert. Using various image processing techniques such as the Convolutional Neural Network (CNN), the Support Vector Machine (SVM), and the k-Nearest Neighbor (k-NN) algorithm, plant diseases can be identified. A user-friendly and robust user interface is developed using web technologies. The fusion and comparison of different image processing algorithms result in a model that is used to predict leaf diseases. The optimal performance in this research is achieved by implementing the CNN+CNN fusion model. This work's unique novel contribution is that it performs all of the machine learning with cloud computing on the AWS EC2 (Amazon Elastic Compute Cloud), whereas previous methods have only stored the data on the cloud; and, the proposed method of this work, fusion CNN + CNN has a 95% accuracy, higher than the other previous methods compared