Change Detection In Multi-Temporal Images Using Multistage Clustering For Disaster Recovery Planning

Change detection analysis on multi-temporal images using various methods have been improved by many researchers in the field of spatial data analysis and image processing. Change detection analysis has many benefit for real-world applications such as medical image analysis, valuable material detector, satellite image analysis, disaster recovery planning, and many others. Indonesia is one of the most countries that encounter natural disaster. The most memorable disaster happened on December 26, 2004. Change detection is one of the important parts of management planning for natural disaster recovery. This article presents the fast and accurate result of change detection on multi-temporal images using multistage clustering. There is three main step for change detection in this article, the first step is to find the image difference of two multi-temporal images between the time before the disaster and after a disaster using operation log ratio between those images. The second step is clustering the difference image using Fuzzy C means divided into three classes. Change, unchanged, and intermediate change region. Afterword the last step is to cluster the change map from fuzzy C means clustering using k means clustering, divided into two classes. Change and unchanged region. Both clustering's based on Euclidian distance. The accuracy achieved using this method is about 98.80%.