Multi-threshold crossover tuna swarm optimization (MT-CTSO) segmentation and residual network (RESNET) classifier for maize leaf diseases

Maize cultivation is frequently affected by diseases, and if they are not promptly prevented and managed, they can lead to reduced yield and quality, resulting in economic losses for farmers. Machine vision technology has been used to identify maize leaf diseases; however, detecting the Region of Interest (RoI) is challenging due to the irregular shapes of the affected areas. Otsu’s threshold-based segmentation offers precise RoI detection. Exhaustive threshold selection significantly degrades performance as the number of thresholds increases, causing exponential computational complexity. To address this, Averaging Histogram Equalization (AVHEQ) is used to denoise disease-affected images, and a novel Crossover Tuna Swarm Optimization (CTSO) algorithm determines the optimal segmentation threshold. Otsu’s method, a threshold-based approach, is also used for accurate RoI detection. In this study, AHEQ is applied to eliminate noise from disease-affected images, while a novel CTSO algorithm is employed to determine optimal segmentation thresholds of objective functions. The outcomes of ten-fold cross-validation were compared with those of previous studies utilizing assessment metrics like Accuracy (Acc), Precision (Pre), Sensitivity (Sen), and Error (Err).