Abrupt change detection on multitemporal remote sensing images: a statistical overview of methodologies applied on real cases Tarek HABIB 1,2 , Jocelyn CHANUSSOT 1 , Jordi INGLADA 2 , Gr´ egoire MERCIER 3 1 GIPSA-Lab (Signal and Images Department), CNRS, INP Grenoble 2 Centre national d’´ etudes spatiales (Cnes) 3 TAMCIC/TIME, CNRS, GET/ENST Bretagne Abstract— In the framework of the International Charter “Space and Major Disasters”, Charter calls are made to the signing parties every time a natural or technological hazard occurs. Consequently, space data are provided by the partners in order to help local authorities to assess the damages, organize and optimize the use of available resources. In such cases, abrupt change detection algorithms are required and numerous methods have been proposed by the Geoscience and Remote Sensing (GRS) community. In this paper statistics measured on Charter calls are compared to statistics measured on change detection methods published in the literature. These statistics aim to give the image processing community a better understanding of the needs and challenges faced in the case of real life disaster scenarios. I. I NTRODUCTION In order to optimize the management of major disasters (flooding, fire, earthquake, tsunami etc.), the use of satellite remote sensing data can be of the utmost importance [1]. In complex geographical zones that are difficult for direct human access and requiring a fast and large scale coverage, remote sensing imagery is a powerful tool for resource management [2]. To ensure a fast and efficient distribution of the data when required, the International Charter ”Space and Major Disasters” – http://www.disastersCharter.org/ – has been signed by numerous administrations and agencies. When a major natural or technological hazard occurs 1 ,a Charter call is made to the signing parties and available data are distributed for an optimal management of the disaster. One of the key issues is then to compare data acquired before and after the disaster and detect abrupt changes [3], [4]. Between January 2001 and May 2006, there have been 96 Charter calls. In this paper, a statistical analysis of these calls is presented to sketch a panorama of the actual needs (type of hazards, type of required products etc.). In the meanwhile, change detection has gathered tremen- dous attention over the past few years and numerous algo- rithms have been proposed by the GRS community. As a matter of fact, over a hundred papers addressing this problem can be found in the recent literature. These algorithms can be classified in to several categories, depending on the type of 1 According to the Charter, natural disasters include earthquakes, tornadoes, cyclones, volcanic eruptions, floods or forest fires, while technological disas- ters include pollution by hydrocarbons, toxic or radioactive substances data, the required supervision, etc. A brief statistical review of change detection methods is presented. The aim of this paper is to put into perspective the needs, as analyzed from the Charter calls, and the available means, as found in the literature. This paper is divided as follows: section II presents statistics measured on the Charter calls, section III presents statistics measured on the change detection methodologies, section IV compares the statistics measured on the methodologies to those of the Charter calls and finally the conclusions are drawn in section V. II. CHARTER CALLS In the period between January 2001 till May 2006, 96 Charter calls were made. This period of five years could be representative of all real life scenarios that could face the inter- national space data community. From floods and earthquakes to hurricanes and tsunamis, this period has seen a multitude of natural and technological hazards. According to this, statistics and measured figures on this period should provide a good perception of challenges that face the actual GRS community. In this section the statistics measured are concerned with the types of hazards, followed by a presentation of the different space products that could be delivered and the relevance of the products to different scenarios is discussed. Six main types of hazards were identified while examining the history of the Charter calls. These types are: floods, earthquakes, forest fire, hurricanes, landslides and technology related hazards. Fig. 1 shows the percentage of occurrence of each type. Flooding events are by far the main source for Charter calls. This has a technical reason, since in the case of floods and due to the nature of the remote sensors, adequate space data could be very rapidly provided to the interested parties. From the study of different Charter calls it was noticed that there is a need to establish norms in order to highlight the most adapted products for each type of hazard, the most suitable image resolutions as well as the maximum time lapse before the delivery of each product. Since these elements are mostly left to personal experience and evaluation, the 1-4244-1212-9/07/$25.00 ©2007 IEEE. 2593