Urban Atlas Updating by Semi-automatic Change Mapping

By the end of 2010, the first edition of the Urban Atlas should be available for over 300 large and medium-sized cities in the EU. This includes European capitals and a number of cities participating in the European Urban Audit – a programme to collect and compare statistical data from some 300 participating cities in Europe. This map will provide, for the first time, homogeneous and up-to-date information on urban environments, allowing for a comparison of different cities all across Europe. Future editions of Urban Atlas are planned in three-to-five-year intervals in parallel with the Urban Audit exercise. The efforts needed for these future editions should be minimized by using an efficient Urban Atlas updating procedure, benefitting from previous efforts.

Within geoland2, one of the tasks addresses the Local Land Monitoring Core Service (i.e. the Urban Atlas) to explore the capabilities for updating these Urban Atlas maps. The produced Urban Atlas data are evaluated to assess its potential for a semi-automatic change mapping approach. Goal is thereby to develop a standard updating protocol guarantying the creation of homogeneous future editions of the Urban Atlas (with a first reference year 2011) and the corresponding change database. The overall aim is to produce an updated Urban Atlas (UA(t1)) database and a database of land cover changes between the first inventory (t0) and the current situation (t1) (UAchanges).

An object-based procedure for detecting infrastructure changes by VHR imagery is proposed. The change analysis procedure consists of different steps: pre-processing, including pan-sharpening, geometric correction and radiometric balancing, object and feature extraction, and statistical change/no-change detection. In a next step the results from these processing steps will be used in a manual on-screen digitizing in order to evaluate and interpret the detected changes. Specific change mapping guidelines have been drawn up to harmonise this visual interpretation. In general, the creation of the updated Urban Atlas can be divided into three steps: (1) an automatic comparison of image data to detect changes – pre-screening; (2) change interpretation and mapping; (3) creation of the 2nd edition map by combining the first inventory of the Urban Atlas with the created change database.

This paper will give an overview of the different methodologies tested in geoland2 applied to land use / land cover change detection in the urban environment. Tested algorithms include image differencing, segmentation comparison and post-classification comparison, all taking into account phenomena / processes as mis-registration, histogram matching, change filtering, image transformations, etc.