Urban areas are important environments where approximately half of the world’s population lives. These centres attract population because they offer greater opportunities for development. Urban sprawl phenomenon is produced due to the fast growing of cities and it entails diverse environmental consequences Therefore, it is necessary to develop technologies and methodologies that permit monitoring the effects of the various problems that are partially caused by urban sprawl. These technologies would help 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 enable the development of more efficient urban and territorial plans. The high dynamism of urban areas produces a continuous alteration of land cover and use, and consequently, cartographic information is quickly outdated. Therefore, the availability of detailed and up-to-date cartographic and geographic information is imperative for an adequate management and planning of urban areas. Usually the process of creating land-use/land-cover maps of urban areas involves field visits and classical photo-interpretation techniques using aerial imagery. These methodologies are expensive, time consuming, and also subjective. Digital image processing techniques help reduce the volume of information that needs to be manually interpreted. These techniques satisfy current demands for continuously precise data that accurately describes a territory.
The aim of this Thesis is to establish a precise methodology to automatically detect buildings and to automatically classify land uses in urban environments using high spatial resolution imagery and LiDAR data. These data are acquired in the framework of the Spanish National Plan for Airborne Ortophotographs, being these data available for public Spanish administrations.
Two main approaches for automatic building detection and localization using high spatial resolution imagery and LiDAR data are adapted. The thresholding-based approach is founded on the establishment of two threshold values: one referred to the minimum height to be considered as building, defined using the LiDAR data, and the other referred to the presence of vegetation, which is defined according to the spectral response. The other approach follows the standard scheme of object-based image classification: segmentation, feature extraction and selection, and classification, here performed using decision trees. In addition, the effect of the inclusion in the building detection process of contextual relations with the shadows is evaluated. Quality assessment is performed at two different levels: area and object. Area-level evaluates the building delineation performance, whereas object-level assesses the accuracy in the spatial location of individual buildings.
Urban land-use classification is achieved by applying object-based classification techniques. Objects are defined using the cartographical limits of cadastral plots. The characterization of the plots to achieve the classification is performed by considering a descriptive feature set, specifically designed to describe urban environments. The proposed descriptive features aim to emulate human cognition by numerically quantifying the properties of the image elements and so enable each to be distinguishable. These features describe each plot as a single entity based on several aspects that reflect the information typology used: spectral, three-dimensional, geometry. In addition, a set of contextual features were defined at two levels: internal and external. Internal context features describe an object with respect to the land cover types contained within the plots, in this case were buildings and vegetation. External context features characterise each object by considering the common properties of adjacent objects that when combined create an aggregation that is higher than plot level: urban blocks.
Results show that thresholding-based building detection approach performs better in the different scenarios analyzed. This method produces a more accurate building delineation and object detection than the object-based classification method. The building type appears as a key factor in the building detection performance. Thus, urban and industrial areas show better accuracies in detection metrics than suburban areas, due to the small size of suburban constructions, combined with the prominent presence of trees in suburban classes, that makes the building detection more difficult. The consideration relations between buildings and shadows improve the object-level detection, removing small objects erroneously detected as buildings that negatively affect to the quality indices.
Classification tests results show that internal and external context features suitably complement the image-derived features, improving the classification accuracy values of urban classes – especially between classes that show similarities in their image-based and three-dimensional features. Context features enable a superior discrimination of suburban building typologies, of planned urban areas and historical areas, and also of planned urban areas and isolated buildings.
These automatic methodologies are especially suitable to compute useful information for constructing and updating land-use/land-cover geo-spatial databases. Digital image processing based methodologies provides better results than visual interpretation based methods. Thus, automatic building detection techniques produce a superior estimation of built-up surface in a objective manner, independently of human operators. The combination of building detection and automatic classification of land uses in urban areas enable to exhaustively distinguish and describe different urban typologies, contributing with more accuracy and information than standard techniques based on visual interpretation. The proposed methodology, based on automated descriptive feature extraction from LiDAR data and images, is applicable for mapping cities, urban landscape characterisation and management, and updating geospatial databases, providing new tools to increase the frequency and efficiency of urban studies.