Image Segmentation of Palm Leaf Pests to Determine Caterpillar Egg Populations Using Marker Watershed

Abstract

Oil palm is an important industrial plant producing cooking oil, industrial oil, and fuel as call biodiesel. One of the factors that can cause a decrease in production yields on oil palm plants is pests. Palm oil companies through the Pest and Plant Diseases team prevent the breeding of pests by taking leaf samples first, for leaf sampling, they must carry out the stages of preparing an early observation schedule, determining sample points and sample lines, and determining sample subjects and it will take quite some time to get the results. Digital image processing is a method for processing digital images and producing other images according to their respective needs so that they are easily interpreted by humans and computers. Detection of caterpillar egg populations found in oil palm leaves is carried out using digital image processing, making it easier for the Pest and Plant Disease team to carry out direct detection of oil palm leaves to determine the caterpillar egg population and find solutions to handle caterpillar eggs. In this study, the population of caterpillar eggs contained in oil palm leaves was detected using the Watershed Marker-based Segmentation method. The stage of image processing begins with taking data obtained from one of the palm oil companies in Riau, then cropping is carried out and followed by color segmentation using Hue Saturation and Value by taking the Value score and then segmenting the marker watershed. The method of testing the credibility of the system uses the one feature method: single decision threshold. The results of testing the credibility of the system obtained a Sensitivity Value Percentage of 90.8% so that there were still 9.2% of the number of caterpillar eggs of oil palm leaf pests that had not been identified and the system accuracy was obtained at 89.4%.