An agent-based method for automatic building recognition from lidar data

• Published in: Canadian journal of remote sensing 2010-06, Vol.36 (3), p.211-223
• Main Authors: Samadzadegan, F., Mahmoudi, F. Tabib, Schenk, T.
• Format: Article
• Language: English
• Subjects: Algorithms; Construction; Information processing; Lidar; Methodology; Multiagent systems; Pattern recognition; Recognition; Remote sensing; Trees; Urban areas
• ISSN: 0703-8992; 1712-7971
• DOI: 10.5589/m10-032

Light detection and ranging (lidar) as a modern and powerful remote sensing technology has proven to be a promising data source for modelling of various three-dimensional (3D) objects such as buildings and trees in urban areas. Nevertheless, because of the adjacency of buildings and other objects, especially trees, the results obtained from most traditional building-recognition algorithms are still dependent on several assumptions and simplifications. This paper presents a multi-agent methodology for automatic building recognition based on the decision-level fusion of textural and spatial information extracted from lidar range and intensity products. In the proposed methodology, two different groups of object-recognition agents are defined for building and tree detection in parallel. The algorithm has two different operational levels based on the types of contextual information. In the first level, both object-recognition agents decide on the types of objects in the study area based on textural information, and the candidates of the building and tree regions are generated. In the second operational level, building- and tree-recognition agents perform some operations at the macro level to modify the candidates of building and tree regions based on spatial information. Evaluation of the results confirms the significant capabilities of the proposed multi-agent algorithm to decrease the conflicts in the field of automatic building recognition in complex urban areas.