Building Detection and Extraction

An automatic building detection technique using LIDAR data and multispectral imagery has been proposed. Two masks are obtained from the LIDAR data: a 'primary building mask' and a 'secondary building mask'. The primary building mask indicates the void areas where the laser does not reach below a certain height threshold. The secondary building mask indicates the filled areas, from where the laser reflects, above the same threshold. Line segments are extracted from around the void areas in the primary building mask. Line segments around trees are removed using the normalized difference vegetation index derived from the orthorectified multispectral images. The initial building positions are obtained based on the remaining line segments. The complete buildings are detected from their initial positions using the two masks and multispectral images in the YIQ colour system. It is experimentally shown that the proposed technique can successfully detect buildings, when assessed in terms of 15 indices including completeness, cor-rectness and quality.

Building detection in complex scenes is a non-trivial exercise due to building shape variability, irregular terrain, shadows, and occlusion by highly dense vegetation. In this research, we present a graph based algorithm, which combines multispectral imagery and airborne LiDAR information to completely delineate the building boundaries in urban and densely vegetated area. In the first phase, LiDAR data is divided into two groups: ground and non-ground data, using ground height from a bare-earth DEM. A mask, known as the primary building mask, is generated from the non-ground LiDAR points where the black region represents the elevated area (buildings and trees), while the white region describes the ground (earth). The second phase begins with the process of Connected Component Analysis (CCA) where the number of objects present in the test scene are identified followed by initial boundary detection and labelling. Additionally, a graph from the connected components is generated, where each black pixel corresponds to a node. An edge of a unit distance is defined between a black pixel and a neighbouring black pixel, if any. An edge does not exist from a black pixel to a neighbouring white pixel, if any. This phenomenon produces a disconnected components graph, where each component represents a prospective building or a dense vegetation (a contiguous block of black pixels from the primary mask). In the third phase, a clustering process clusters the segmented lines, extracted from multispectral imagery, around the graph components, if possible. In the fourth step, NDVI, image entropy, and LiDAR data are utilised to discriminate between vegetation, buildings, and isolated building's occluded parts. Finally, the initially extracted building boundary is extended pixel-wise using NDVI, entropy, and LiDAR data to completely delineate the building and to maximise the boundary reach towards building edges. The proposed technique is evaluated using two Australian data sets: Aitkenvale and Hervey Bay, for object-based and pixel-based completeness, correctness, and quality. The proposed technique detects buildings larger than 50 m 2 and 10 m 2 in the Aitkenvale site with 100% and 91% accuracy, respectively, while in the Hervey Bay site it performs better with 100% accuracy for buildings larger than 10 m 2 in area.

Automatic 3D extraction of building roofs from remotely sensed data is important for many applications including city modelling. This paper proposes a new method for automatic 3D roof extraction through an effective integration of LIDAR (Light Detection And Ranging) data and multispectral or-thoimagery. Using the ground height from a DEM (Digital Elevation Model), the raw LIDAR points are separated into two groups. The first group contains the ground points that are exploited to constitute a 'ground mask'. The second group contains the non-ground points which are segmented using an innovative image line guided segmentation technique to extract the roof planes. The image lines are extracted from the grey-scale version of the orthoimage and then classified into several classes such as 'ground', 'tree', 'roof edge' and 'roof ridge' using the ground mask and colour and texture information from the orthoimagery. During segmentation of the non-ground LIDAR points, the lines from the latter two classes are used as baselines to locate the nearby LIDAR points of the neighbouring planes. For each plane a robust seed region is thereby defined using the nearby non-ground LIDAR points of a baseline and this region is iteratively grown to extract the complete roof plane. Finally, a newly proposed rule-based procedure is applied to remove planes constructed on trees. Experimental results show

Despite the large number of studies conducted during the last three decades concerning 3D building modelling starting from Lidar data, two persistent problems still exist. The first one is the absence of some roof details which will not only disappear in the building roof model due to their small areas regarding the point density, but are also considered as undesirable noise among the modelling procedures. The second problem consists in that the involved segmentation algorithms do not perform well in the presence of noise in the building point cloud data. These two problems generate undesirable deformation in the final 3D building model. This paper proposes a new automatic approach for detecting and modelling the missing roof details in addition to improving the building roof segments. In this context, the error map matrix, which presents the deviations of points to their fitting planes, is considered. Moreover, this matrix is analyzed in order to deduce the mask of missing roof details. At this stage, a new numeric factor is defined for estimating the roof segmentation accuracy in addition to the validity of the roof segmentation result. Then, the building point cloud is enhanced in order to decrease the negative noise influence and, consequently, to improve the building roof segments. Finally, the functionality and the accuracy of the proposed approach are tested and discussed.

Three-dimensional (3-D) reconstruction of building roofs can be an essential prerequisite for 3-D building change detection, which is important for detection of informal buildings or extensions and for update of 3-D map database. However, automatic 3-D roof reconstruction from the remote sensing data is still in its development stage for a number of reasons. For instance, there are difficulties in determining the neighbourhood relationships among the planes on a complex building roof, locating the step edges from point cloud data often requires additional information or may impose constraints, and missing roof planes attract human interaction and often produces high reconstruction errors. This research introduces a new 3-D roof reconstruction technique that constructs an adjacency matrix to define the topological relationships among the roof planes. It identifies any missing planes through an analysis using the 3-D plane intersection lines between adjacent planes. Then, it generates new planes to fill gaps of missing planes. Finally, it obtains complete building models through insertion of approximate wall planes and building floor. The reported research in this paper then uses the generated building models to detect 3-D changes in buildings. Plane connections between neighbouring planes are first defined to establish relationships between neighbouring planes. Then, each building in the reference and test model sets is represented using a graph data structure. Finally, the height intensity images, and if required the graph representations, of the reference and test models are directly compared to find and categorise 3-D changes into five groups: new, unchanged, demolished, modified and partially-modified planes. Experimental results on two Australian datasets show high object- and pixel-based accuracy in terms of completeness, correctness, and quality for both 3-D roof reconstruction and change detection techniques. The proposed change detection technique is robust to various changes including addition of a new veranda to or removal of an existing veranda from a building and increase of the height of a building.

—Some performance evaluation systems for building extraction techniques are manual in the sense that only visual results are provided or human judgement is employed. Many evaluation systems that employ one or more thresholds to ascertain whether an extracted building or roof plane is correct are subjective and cannot be applied in general. There are only a small number of automatic and threshold-free evaluation systems, but these do not necessarily consider all special cases, for example, when over-and under-segmentation occurs during the extraction of roof planes. This paper proposes an automatic and threshold-free evaluation system that offers robust object-based evaluation of building extraction techniques. It makes one-to-one correspondences between extracted and reference entities using the maximum overlaps. Its application to the evaluation of a building extraction technique shows that it estimates different performance indicators including segmentation errors. Consequently , it can be employed for bias-free evaluation of other techniques whose outputs consist of polygonal entities.