Automated 3D Modeling of Buildings in Suburban Areas Based on Integration of Image and Height Data

Abstract

This paper presents an automated method for 3D modeling of buildings in suburban areas through the integration of image and height data. The method is based on matching a CAD model of the building against the image, while the selection of the CAD model relies on clues derived from height data. The matching procedure makes use of straight line segments extracted from the image and grouped on the basis of proximity and parallelism relations. For the selection of the model, a process of fitting planar faces to height data guided by a segmentation of image data is employed. Roof planes are recognized by taking into account the height of each plane over a DTM of the scene. The integration strategy proposed in this paper is capable of exploiting accurate and reliable information from both sources of data. Incomplete image regions can be refined by using clues from height data, and incomplete image edges can be efficiently handled in the model matching process. The reconstruction strategy takes advantage of the high accuracy of laser range data, but is not influenced by the low spatial resolution of the DSM. An experiment is conducted to evaluate the performance of the proposed approach. Results indicate the promising performance of the proposed approach in reconstructing buildings with an acceptable accuracy at a reasonable computational cost.