Automatic Assessment of Earthquake Damaged Area Using Scale- Space Classification Techniques

Disaster management has got an important facet, viz., disaster mitigation, which in turn depends upon early damage assessment. Remote Sensing is considered as an ideal technique for obtaining early information. This exercise has been carried out for assessing the damaged area, automatically, based on high-resolution post-earthquake images in Gujarat, India and high resolution satellite images of tsunami-affected areas in India and Sri Lanka. The damaged area in the form of rubble manifested as fine structure in the image. The separation of the fine structure from the rest of the image is attempted by scale-space representation approach. This method is found to be very effective for estimation of damaged area automatically in quick time and can be a very useful tool in the disaster management. 1.0 Introduction: In recent years Disasters like Earthquakes and Tsunami’s have caused thousands of deaths and severe economic losses. The impact of theses disasters are many and varied, ranging from minor structural damages to few buildings to complete annihilation of major urban centers and severe disruption of the social and economic fabric of the nation. In 1994 the Northridge quake in Southern California and the Kobe earthquake in Japan 1995 resulted in thousands of deaths and economic losses estimated at around 200 billion US Dollars. Last Gujarat earthquake in India resulted in approximately 20,000 deaths, 1,66,000 injured with 3,48,000 houses destroyed and 8,44,000 houses damaged. The Indian government estimated that the earthquake affected, directly or indirectly, 159 lakhs people and direct economic losses of 6000 crores of rupees. The devastating tsunami in Indian Ocean in the year 2004, killed more than 2,25,000 people in eleven countries and caused heavy economic losses. Disaster mitigation is needed to reduce worsening of postdisaster situation by way of taking effective steps to carry out relief work, finding suitable relocation places, etc. Quick information of the damaged areas (badly hit area) is very essential to do the relief work immediately so that the assistance can reach the needy people at the earliest. In the event of disaster like Earthquake and Tsunami, conventionally, first hand information about the damage affected areas can be obtained earliest through aerial surveys due to obvious reasons. Now with the availability of multiple high-resolution satellite sensors like IKONOS, Quickbird etc. and increasing frequency of coverage satellite remote sensing can be considered as a viable source of early information. The availability of high-resolution satellite/aerial imagery will make a profound contribution to earthquake damage assessment and disaster mitigation. Quite often it is realized, although data can be acquired reasonably quickly, it takes far longer to process/analyze and create spatial information of relevance due to the large amount of data to be analyzed manually. Manual processing of data becomes so time consuming, tedious and it varies individually. To make the process faster, simpler and robust we need to have an automatic method, which gives us the knowledge of the areas damaged by these disasters. Damaged area assessment can be easily carried out when preand post-event images are available. It can be assessed, using simple change detection techniques. But events like earthquakes and tsunamis are usually unpredictable in nature hence pre event data may not be available. In this paper we are using a multi-scale approach to assess the damaged area with the help of a single post-disaster image. From the image point of view the signature of damages was found to be the similar for the earthquake and Tsunami disaster. Hence the same method is applicable in both the cases. 2.1 Scale-Space theory: Scale is inherently tied up with information. Optimal information extraction can be carried out at certain scales. For example extraction of 20-meter wide road by 1-meter resolution image data will show not only the road but also features like dividers, markers, vehicles, shadows, etc. making the road extraction a complex process. At the same time if the scale of the image is reduced, i.e., effectively finer structures/information is eliminated then the problem becomes simpler. By the same reason, some of the features will appear over certain range of scale. Scale Space theory provides a welldefined framework for dealing with image structures at different scales [Lindeberg, 1994]. The basic idea behind a multi-scale representation of signal is by generating one-parameter family of derived signals, where fine-scale information is successively suppressed. This operation, which will be termed scale-space smoothing, must be available at any level of scale. The major reason for this is to