BUILT UP CHANGE DETECTION BASED ON STRUCTURAL CLASSIFICATION OF DIGITAL ELEVATION MODELS
Keywords: Buildings, Normalized DSM, Segmentation, Structural Classification, Change Detection
Abstract. Multi temporal changes in built up areas are mainly caused by natural disasters (such as floods and earthquakes) or urban sprawl. Detecting these changes which consist of construction, destruction and renovation of buildings can play an important role in updating three dimensional city models and making the right decisions for urban management. Generally, change detection methods based on multi temporal remotely sensed data can be divided into 2D and 3D categories. Three dimensional change detection methods are suitable for identifying the changes of three dimensional objects such as buildings and their results are more close to reality. The objective of this study is to provide an effective method for 3D change detection of buildings in urban areas based on Digital Elevation Models (DEM). The proposed method in this paper consists of three main steps; 1) generating the normalized DSM for two epochs, 2) performing segmentation and structural classification of image segments in order to generate multi temporal classification maps, 3) producing the change maps. The ability of the proposed algorithm is evaluated in a rapid developing urban area in Tehran, Iran in a 9-years interval. The obtained results represent that the ground and bare soil decrease for about −1.37% and low-rise buildings also decrease for about −9.7%. Moreover, the class of high-rise buildings increases for about +16.4% which conforms making new constructions in addition to renovation of low-rise buildings.