Building Classification

Object recognition and image classification has become an area of intense focus in machine learning and pattern recognition. Building recognition and classification rise into attention in the area of urban reconstruction where a number of images of buildings are used. We first compare different classification approaches to discriminate the class building from the class non-building. The images are obtained from the Caltech256 database. Secondly we recognize individual buildings in the ZuBuD database, consisting of 201 individual buildings. We compare three different algorithms: Consistent Line Clusters(CLC), Randomized Decision Trees and the Vocabulary Tree. We obtain 99% accuracy on the building versus non-building classification and 75.6% on specific building classification. We conclude that the CLC method performs best in first case, while the Vocabulary Tree performs best in the second case, although Randomized Trees perform similar (72%).