Design of a real-time embedded system that detects diseases on tobacco leaves using morphological image processing methods

Since it does not have very rigorous conditions of cultivation, the tobacco plant is produced in all regions of Turkey, except for the Central Anatolia region. The tobacco plant, which has huge economic importance, has reached an export figure of $828.9 million in 2022. Diseases emerging in tobacco leaves act against the normal and healthy growth of the tobacco plant. This case causes yield loss and thus financial losses. This detection of diseases in tobacco leaves does cause time loss and could also be inaccurate. So a real-time embedded system model has been designed to detect the area of tobacco leaves that are diseased. In all these systems, images taken from the camera are transferred to the embedded system. After that comes the application of morphological image processing, with the help of a Python software, which allows the detection and display of the areas of the leaves affected by such disease. Classification of tobacco leaves into categories such as diseased and healthy was also done for developing convolutional neural network models. Performance was tested on a dataset of 1600 healthy and 1600 diseased tobacco leaves pictured in Samsun province. Through the classification process, it succeeded in success rates of 93% or more in individual three models. Keywords : Tobacco leaf diseases, Embedded systems, Convolutional neural networks, Classification