Abstract
Abstract. We present a study for the evaluation of the efficiency of context features in object-based land-use classification of urban environments using aerial high spatial resolution imagery and LiDAR data. Objects were defined by means of cartographic boundaries derived from the cadastral geospatial database. Objects are exhaustively described through different types of image derived features (i.e. spectral and texture), three-dimensional features computed from LiDAR data, and geometrical features describing the shape of each object. Additionally, the context of each object is described considering several aspects: adjacency, urban morphology, vegetation, and geometry. Adjacency between objects was characterized by features computed using the graph theory. Urban morphology features are related to the shape and size of neighbouring buildings, and are often related to their socio- economic function. The presence and density of vegetation are strongly related to the different urban typologies. Many of the contextual features are related to buildings, which are obtained by means of automatic building detection techniques. The meaning of the defined features, and their contribution to the classification accuracy were analyzed. The results showed that the inclusion of contextual features had a positive effect on land use classification of urban environments, increasing the overall accuracy around 4%, compared of using only the rest of features. The classification efficiency particularly increased in some classes, such as different typologies of suburban buildings, planned urban areas and historical areas.
Highlights
Half of the world’s population is currently living in cities and this proportion is expected to increase progressively to 70% by 2050 (United Nations, 2010)
To visual techniques used by photointerpreters, digital image processing techniques describe urban elements properties through image derived features, three-dimensional features computed from LiDAR data, geometrical features describing the shape of each object, and contextual features which are related to the spatial attributes of the overall environment
Objects were exhaustively described through image derived features, three-dimensional features computed from LiDAR data, and geometrical features relating the shape of each object
Summary
Half of the world’s population is currently living in cities and this proportion is expected to increase progressively to 70% by 2050 (United Nations, 2010). The global increase in urban population and the rapid urbanisation processes was first experienced in developed countries in the middle of the twentieth century, and it is currently occurring in developing countries of Africa, Asia and Latin America. Urban sprawl phenomenon is produced due to the fast growing of cities and it entails diverse environmental consequences such as increasing the dependence on cars. It is necessary to develop technologies and methodologies for monitoring urban sprawl and the side effects it causes. Sensed data would enable the rapid adoption of policies that minimise the negative effects of urban sprawl. Solutions require a precise knowledge of the current urban environment to develop more efficient urban and territorial plans
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
