HSO Based K-Means for Automatic Segmentation of Disease and Deficiency in Tomato Plant

Abstract

Image segmentation is an important part in image processing because it split the given whole image into parts, among the parts we can choose which one is most important. There are wide range of image processing algorithms available among them K-means algorithm is very simple to understand is one of the simplest and Produce accurate result, even though it is simple and accurate it has some limitations that is we need to guess the value for K and randomly select K initial centroids among the given data points. To overcome this initialization problem a new method for image segmentation based on Harmonic search optimization (HSO) and K-means for deficiency detection was proposed. The performance of K-means algorithm is mainly based on the k value and k initial centroids of the clusters. Random initialization is followed in normal K-means algorithm due to this normal k-means take lots of time to produce correct. A new method called HSO based K-means is proposed to speed up the initialization process. The proposed algorithms exploit an initial step derived from the HSO, considering Otsu method as the objective function. After finding the cluster centers using HSO, K-means algorithm is initialised with these cluster centers. Finally, segmentation result is compared with normal K-Means and EM segmentation algorithm our proposed HSO based K-means algorithm gives better result than others.