Detection of Tsunami Induced Changes from High Resolution Satellite Imagery

Abstract

The 2004 Indian Ocean earthquake, also known as the Sumatra-Andaman earthquake, was an undersea earthquake that occurred at 00:58:53 UTC (07:58:53 local time) on December 26, 2004. The tsunami, generated by the earthquake killed approximately 275,000 people, injured lots more and forced many individuals to leave their homes. The earthquake originated in the Indian Ocean just north of Simeulue Island, off the western coast of northern Sumatra, Indonesia. The resulting tsunami devastated the shores of Indonesia, Sri Lanka, South India, Thailand and other countries with waves up to 30 m (100 ft). In this study, the detection of changes due to the tsunami after the earthquake is performed. The images of the suffered region are acquired by Quick Bird satellite of Digital Globe Company, taken before (on April 12, 2004) and after (on January 2, 2005) the disaster. The aim of this study is to find the location of the changed regions and to specify the intensity of the change by using various change detection algorithms, which are the image algebra (band differencing and band rationing), post-classification comparison, the binary mask applied to date-2 and write function memory insertion. The image algebra techniques are generally used in two forms, which are called band differencing and band rationing. Band differencing is probably the most widely applied change detection algorithm that involves subtracting one date of imagery from a second date that has been precisely registered to the first. The resulting image contains both negative and positive values, which indicate the change between the images. Intuitively, if there is no change, then the expected values from image differencing would be zero. Similar to image differencing, change detection can also be achieved through band rationing. The basic idea is to create ratios between two different images of the same area. The areas of no change in this procedure will result in value of 1, where changes greater and less than 1 indicates the differences between the images. In post classification comparison, each image is classified by using a supervised classification algorithm (e.g. Maximum Likelihood Classifier) and the classified images are compared pixel by pixel In the binary mask applied to date-2 method, the pre-event image is firstly classified and then a new image is generated by performing algebraic operations on an arbitrary common band of the pre-and post-event images. In addition to that, a threshold value is determined whether a change occurs or not. The regions under the threshold are removed and the remaining parts are specified as changed regions. In the last technique, write function memory insertion, the individual bands of the images are inserted into specific write function memory banks (red, green or blue) in a digital image processing system. During the implementation of these methods, Matlab programming language, which is quite efficient in image processing operations, is used. The results of this study indicate the regions that are changed by the tsunami and the intensity of this change are successfully detected.