Harumanis Mango Leaves Image Segmentation on RGB and HSV Colour Spaces using Fast k-Means Clustering

Abstract

Harumanis mango is the signature fruit in Perlis due to its delicious taste and its sweet-smelling. A good quality Harumanis tree requires rich in nutrition (healthy), and the tree will grow lots of fruits compared to the trees which are poor in nutrition (unhealthy). The health condition of a tree can be observed through the leaves in term of shape of leaves. For a healthy Harumanis tree, the leaves grow in scattering shapes. Meanwhile, an unhealthy Harumanis tree grows in gathered shapes. Therefore, this research is focusing on Harumanis mango leaves image segmentation by comparing between RGB and HSV colour spaces in order to obtain the best segmentation performance. 100 of Harumanis mango tree leaves images are used in this research. These images have undergo through image pre-processing such as modified linear contrast stretching and colour components extraction based on RGB and HSV colour spaces. Then, the colour component images have been segmented by using fast k-means clustering in order to obtain the leaves segmented images. Finally, quantitative analyses have been performed to measure the segmentation performance based on sensitivity, specificity and accuracy. Overall, the results show that S component of HSV colour space archives the highest accuracy with 85.81%.