Change Detection for Aerial Photo Database Update

With rapid new technological development during the past 50 years, aerial photos have been increasingly and commonly used not only in geographic information systems but also in various spatially related applications. Many municipalities and government agencies have constructed aerial photo databases all over the world. Keeping these databases up to date is the most important part of making them effective so that aerial photo databases are expected to be updated as frequently as possible. However, in practice, some of them are barely updated because of high cost. In case of detecting where changes occur, aerial photo databases could be frequently and partially updated only in the changed areas.In this study, medium spatial resolution imagery is proposed to detect changes. Instead of using aerial photos, medium spatial resolution imagery could be a good alternative because of its wider coverage and lower price than aerial photos. Free accessible Landsat ETM+ and orthophotomaps are used for change detection. In order to compare with 15 m pan-sharpened Landsat ETM+, 1 m orthophotomaps are decomposed, segmented, and classified through wavelet transform and object-oriented classification. Although the detected changes are rough, the result shows that the method is quite costeffective and practical. Moreover, it could support decision making for updating aerial photo databases. * Corresponding author 1. INTRODUCTIOIN Aerial photos have been proved to be useful for extracting spatial information or performing spatial analysis since they have been widely used for mapping purposes. In addition, diverse aerial photo products such as topographic maps, digital elevation models, and orthophotos play an important role in various spatially related applications. These are reasons why many municipalities and government agencies have constructed aerial photo databases in Canada and also in other nations. In Canada, for example, in federal case, the National Airphoto Library (NAPL) by Natural Resources Canada archives over six million aerial photos covering all of Canada. In provincial case, the Alberta Sustainable Resource Development collects over 1.4 million aerial photos covering whole province, and the Provincial Softcopy Orthophotomap Database (SODB) by the Service New Brunswick (SNB) was created from aerial photos for the whole province. Considerable amounts of aerial photo databases have been built all over the country. In general, most databases have an updating problem and it has been a continuous issue in this field. Aerial photo database is no exception, and up-to-date data is indeed necessary to extract and analyze trustworthy spatial information. In practice, the SNB has received numerous inquiries into when the SODB will be updated. Although most aerial photo databases are expected to be updated as frequently as possible, some are barely updated because of high cost. Human activities and natural phenomena would cause landcover changes, and in practice they are not occurred in whole area. Hence, in case of cost-effectively detecting where changes occur, aerial photo databases could be frequently and partially updated only in the changed areas. Multitemporal images from same sensor are generally used in change detection, but it is too expensive to use aerial photos. Medium spatial resolution imagery, such as Landsat and SPOT usually has wider coverage and lower price than aerial photos, so it could be a good alternative for change detection. In this study, a cost-effective and practical method is suggested to detect land-cover changes to support decision making for updating aerial photo databases. It involves wavelet transform and object-oriented classification to compare two different types of images, and then post-classification comparison is applied to find changes.