Evaluation of segmentation methods for RGB colour image-based detection of Fusarium infection in corn grains using support vector machine (SVM) and pre-trained convolution neural network (CNN)

This study evaluated six segmentation methods (clustering, flood-fill, graph-cut, colour-thresholding, watershed, and Otsu’s-thresholding) for accuracy classification in discriminating Fusarium infected corn grains using RGB colour images. The was calculated Jaccard similarity index Dice coefficient comparison with the gold standard (manual method). Flood-fill graph-cut showed highest of 77% 87% evaluation metrics, respectively. Pre-trained convolution neural network (CNN) support vector machine (SVM) were used to evaluate effect on segmented images extracted features from images, SVM based two-class model discriminate healthy yielded 84%, 79%, 78%, 74%, 69% 65% clustering, Otsu’s-thresholding, In pretrained CNN model, accuracies 93%, 88%, 87%, 61% 59% metrics correlation R2 values 0.9693 0.9727, R2–0.505 R2–0.5151 metrics.