Object-based Class Modeling for Cadastre-constrained Delineation of Geo-objects

Abstract

Remote sensing technology still faces challenges when it comes to monitoring tasks that must be able to stand up to validation from technical, scientific, and practical points of view, in other words, when entering into established, fully operational workflows. In this paper, we present an approach for delineating and monitoring aggregated spatial units relevant to regional planning tasks, which has been fully validated within a 3,654 km 2 area in the Stuttgart Region of southwestern Germany. This has been achieved by developing algorithms for semi-automated (geo-) object-based class modeling of biotope complexes, which are aggregated, functionally homogenous (but not necessarily spectrally homogeneous) units. High levels of complexity in the target classes and the need for integration of auxiliary geodata as a priori knowledge meant that different methods of information extraction were required to be combined in an operational workflow, and that new validation strategies were needed for quality assessment. A total of 31,698 biotope complexes were delineated for the entire Stuttgart Region, with an average size of 11.5 ha for each complex. Approximately 86 percent of the biotope complex boundaries were shown to have been correctly delineated.