Embedded system design for real-time detection of tobacco blue mold disease

The tobacco plant is grown in several regions globally as well as Turkey due to its significant adaptability. The occurrence of blue mold disease on tobacco leaves adversely affects the growth and development of the plant, leading to yield and economic losses. The traditional diagnosis of blue mold disease (Peronospora tabacina Adam) in tobacco leaves is time-consuming, which may delay control measures and accelerate the spread of the disease. This situation complicates early and accurate intervention strategies. Therefore, a real-time embedded system model was designed to detect diseased areas on tobacco leaves. Camera images were transferred to the embedded system, and symptomatic regions were identified using morphological operations implemented through Python software. In addition, convolutional neural network (CNN) models were employed to classify tobacco leaves as healthy or diseased. The performance of these models was evaluated on a dataset consisting of 1 600 healthy and 1 600 diseased tobacco leaf images taken in the Bafra district, Samsun, Turkey. As a result of the classification process, the system achieved a success rate of >93% across three different models. The developed real-time embedded system is expected to contribute to preserving productivity and sustainability in agriculture by enabling accurate and rapid detection of blue mold disease in tobacco leaves.