Pre-processing Technique using Colour-based Feature Method to Detect Categories of Leaves Disease

Abstract

Oil palm leaves diseases is caused by various plant pathogens and micronutrient deficiency, and genetic disorders. This problem, if not identified and treated quickly could lead to losses in yield and profitability. The disease on leaves is currently being identified through the different colours, shapes, and forms. Other signs of an infected plant can be seen based on the discolouration on the leaves. In this paper, we present an approach to automatically identify the morphological features of leave diseases in category of healthy to non-healthy based on region of interest of discolouration on young oil palm leaves. Raw leaf images are captured using a built-in digital camera. Pre-processing was done on each of the non-uniform illumination condition raw data images. We tested the colour feature method using RGB (Red, Green Blue) colour filtering in the identification of the leaf region of interest. Next, further segmentation method using HSV (Hue, Saturation, Values) colour filtering approach is employed to remove shadows and to identify the different level of regions of discolouration. The results highlighted that the infected area on the leaves can be identified by 100% based on the discoloured in the region of interest. These regions can be categorised in three different groups – healthy leaves (20% of the discolouration region) to heavily infected (70% of the discolouration region) of the leaves – based on analysis of the pre-processing results. In top of that, the HSV colour feature method could also remove shadow and noise. The results of the detected discolouration will be used oil palm leaves datasets for further classification and recognition research work.