K-mean Clustering for Segmentation of Irregular Shape Fruit Images under Various Illumination

Abstract

Segmentation is the first step in analyzing or interpreting an image automatically. In particular applications, like image compression or image recognition, entire image can�t be processed directly. Hence many segmentation techniques are proposed to segment an image before processing it. This made it possible to develop many techniques which are currently using in different industries and agriculture field. They are either applied for grading or inspecting quality of food products and Fruits. These developed techniques use thresholding and clustering approach to get proper segmented output. In this paper an image segmentation approach is developed based on k-means adaptive clustering. This approach segments the various shape fruit images particularly which are non-circular (like banana, mango, and pineapple) and captured in various illumination such as low, Medium and high intensity. Earlier segmentation methods were not apposite for fruit images captured in natural light; as they were responsive to various colour intensity predisposed by the sunlight illumination. Natural illumination tempts an uneven amount of light intensity on the surface of the object, resulting in poor quality image segmentation. This approach will deal with problem of light effect. K-means clustering is renowned method for image segmentation. This method is more efficient, robust than the others. It provides best result when dataset is well separated and distinct. Different shape fruit images are segmented properly along with grey scale. The analytical results are the evidence for the accurate segmentation of banana, mango pineapple using new approach developed here.