Classification Of Merapi Volcano Images Based on HSV Color Feature Extraction and Local Binary Pattern Texture Feature Extraction Using The K-Nearest Neighbors Method

Abstract

The BPPTKG (Center for Volcanology and Geological Hazard Mitigation) routinely monitors  Merapi Volcano’s activity through visual imagery captured with DSLR lenses at several observation posts. However, not all recorded imagery can be used for analysis due to frequent cloud or fog cover. This not only makes it difficult for experts to accurately monitor Merapi's condition but also reduces the efficiency of data storage capacity. To examine the application of HSV color feature extraction, LBP texture feature extraction, and the K-Nearest Neighbor method for classifying Merapi Volcano images based on appearance. The dataset used consists of Merapi Volcano images captured from the Tunggularum observation post between October 1st and 10th, 2023, categorized into six classes based on the volcano's appearance. Preprocessing steps include cropping, masking, and image sharpening. Classification was performed using the K-Nearest Neighbor method to obtain the classification results of Mount Merapi images. Based our result, the classification method using HSV and LBP using the K-Nearest Neighbor method was successfully performed. The optimal value of k was 1, achieving an accuracy of 95%, while the worst value of k was 9, with an accuracy of 87%.