Determination and classification of diseases in plants with deep learning methods

Due to the rapid developments in the latest technology, agricultural studies have been positively affected by these developments and the concept of sensitive agriculture has emerged. Precision agriculture aims to minimize the human labor in the field and to maximize productivity by using technology beside this. Precision agriculture practices are developing in line with the needs of the farmer. One of the biggest needs of the farmer is that the product farmer's obtains is of high quality and efficient. Therefore, the plant should be kept away from diseases during its life stages. A plant should be observed at every step of the life stage and kept as far away from diseases as possible. Diseases should be detected at an early stage if the plant catches any disease, so that it can be prevented without causing too much damage to the plant and its crop. Detection and rehabilitation of plant disease in the first case may prevent deterioration of the quality of plant products. It is possible to detect the diseases seen in plants automatically and to determine the disease type with precision agriculture applications. In this thesis, detection and classification of diseases in tomato and apple leaves were carried out by deep learning methods. An approach based on Convolutional Neural Network (CNN) model which is one of the deep learning methods and Learning Vector Quantization (LVQ) algorithm is presented. Experimental studies were done separately for tomato and apple leaves. In the data set consisting of tomato leaves, experimental studies were conducted for 5 different classes, 4 different diseased leaves and 1 healthy leaf. In the studies carried out on apple leaves, the images of the leaves in the apple orchards in Yalova province were taken with camera and the experimental studies were carried out on the leaf images on the tree. Experimental results show that the convolutional neural network based method proposed in the study is successful in detecting diseases in leaves and can be used effectively for disease detection and classification studies in agricultural areas.