A Comparative Study among Clustering Techniques for Leaf Segmentation in Rosette Plants

Abstract

Plant image analysis plays one of the key roles in agriculture. Using it, the traits of the morphological plants can be recorded frequently and accurately. One of the important traits to be analyzed is the growth of plants, which deeply rely on the leaves and their segmented images. In this paper, four of the mostly used clustering algorithms are compared by segmenting the leaves of the rosette plants, with an ideal theoretical segmentation, to see how the results differ and which one tends to create the most accurate segment. The four clustering algorithms are namely K-means (KM), particle swarm optimization (PSO), fuzzy C-means (FCM), and self-organizing map (SOM), by which each single image of utilized dataset has been tested and numerical analysis has been put afterwards, which clearly points out that SOM provides the nearest result to the ideal ground truth; hence, the most efficient one.