The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XL-1/W5
https://doi.org/10.5194/isprsarchives-XL-1-W5-601-2015
https://doi.org/10.5194/isprsarchives-XL-1-W5-601-2015
11 Dec 2015
 | 11 Dec 2015

AUTOMATIC BLOCKED ROADS ASSESSMENT AFTER EARTHQUAKE USING HIGH RESOLUTION SATELLITE IMAGERY

H. Rastiveis, E. Hosseini-Zirdoo, and F. Eslamizade

Keywords: High resolution Satellite Image, Earthquake, Classification, Texture Analysis, Roads damage map

Abstract. In 2010, an earthquake in the city of Port-au-Prince, Haiti, happened quite by chance an accident and killed over 300000 people. According to historical data such an earthquake has not occurred in the area. Unpredictability of earthquakes has necessitated the need for comprehensive mitigation efforts to minimize deaths and injuries. Blocked roads, caused by debris of destroyed buildings, may increase the difficulty of rescue activities. In this case, a damage map, which specifies blocked and unblocked roads, can be definitely helpful for a rescue team.

In this paper, a novel method for providing destruction map based on pre-event vector map and high resolution world view II satellite images after earthquake, is presented. For this purpose, firstly in pre-processing step, image quality improvement and co-coordination of image and map are performed. Then, after extraction of texture descriptor from the image after quake and SVM classification, different terrains are detected in the image. Finally, considering the classification results, specifically objects belong to “debris” class, damage analysis are performed to estimate the damage percentage. In this case, in addition to the area objects in the “debris” class their shape should also be counted. The aforementioned process are performed on all the roads in the road layer.In this research, pre-event digital vector map and post-event high resolution satellite image, acquired by Worldview-2, of the city of Port-au-Prince, Haiti's capital, were used to evaluate the proposed method. The algorithm was executed on 1200×800 m2 of the data set, including 60 roads, and all the roads were labelled correctly. The visual examination have authenticated the abilities of this method for damage assessment of urban roads network after an earthquake.