Swarm Intelligence for Segmentation of Leaf Images

Leaf segmentation is important in assisting environmentalists to automatically segment the foreground leaf from the noisy background. The accuracy with which the image is segmented and the unwanted background areas are removed determines the result acquired from machine learning algorithms employed in feature extraction or classification. In this paper, the most commonly used unsupervised clustering algorithm K -means has been discussed, which has further been optimized using particle swarm optimization and firefly algorithms, and the performance of the three techniques has been compared and results have been presented. The evaluation metrics used are sensitivity, specificity, segmentation accuracy, precision, dice, Jaccard distance and MCC. The experiments have been performed on 50 images from each class taken from the PlantVillage dataset.