A Hybrid Metaheuristic Algorithm for Diseases Classification Using UAV Images

Recent advances in technology are very astounding since they have made it possible to manage and monitor systems remotely. Traditional farming is undergoing a transition towards "smart farming" with the assistance of technological breakthroughs, which include the implementation of intelligent irrigation systems and the remote monitoring of the development of crops. In particular, the Unmanned Aerial Vehicle plays a significant role in sophisticated UAVs' ability to capture photographs of crops and spray for pests. The image that is obtained from UAVs is then subjected to various forms of computer-assisted processing in order to determine whether or not the crop's leaves are naturally healthy, diseased, or rotten. Several groups of researchers investigated a variety of approaches, including clustering, machine learning, and deep learning, with the goal of determining the nature of the leaves and categorizing them according to the characteristics they possessed. These traits are necessary for categorization, but the time required to process them will be increased because of their enormous size. Because of this, the authors of this study present a hybrid feature reduction technique that is a blend of two different metaheuristic algorithms. In this case, an upgraded version of the cuckoo search algorithm was paired with the particle swarm to find the most advantageous characteristics. In this, the optimum features of the texture, such as its GLCM, GLDM, and local binary pattern features, were chosen for selection. Using a neural network that was based on the back propagation technique, the optimal characteristics were used for classification. The method that has been suggested is based on photographs that were taken in natural settings of sets of healthy and diseased leaf specimens. The entire process is carried out with the assistance of the MATLAB R2021a program and the results are analyzed with Accuracy, Sensitivity, and Specificity.