Advanced search
Start date
Betweenand

Automated building boundary extraction using very high-resolution aerial images and laser scanning data

Abstract

The automated building extraction from remotely captured data (specifically optical images and laser scanning data) is a complex task, due to e.g. the inherent complexity of such objects and obstructions that exists in the real scene (caused e.g. by vegetation and shadow) or caused by the geometry of the employed sensor. Usually, the three-dimensional building reconstruction process from images or laser scanning data involves the following sequential procedures: boundary detection/delineation (or extraction); roof reconstruction; and the complete three-dimensional reconstruction of the building. Other detection or reconstruction strategies are necessary when combining both data source (laser scanning data and satellite or aerial images). In general, laser scanning data allow the extraction of roof planes and lines of ridge with better quality when compared to results obtained from images. Moreover, it is usually easier to detect/delineate building in laser scanning data than in images. On the other hand, high-resolution images allow the extraction of building contours with superior quality. Therefore, in order to take advantage of the synergy between image and laser scanning data, it is recommended that the aforementioned steps of building extraction should be combined in some way. The specific problem of building boundary extraction is related to the most difficult part of the general problem of three-dimensional extraction of building models. The basic reason is because the building recognition occurs in this step, prior to the building boundary delineation. The combination of building boundaries previously extracted from airborne laser scanning (ALS) and very high-resolution aerial images aims at exploring the fact that it is easier to detect buildings in ALS data and that the ALS building boundaries is helpful for localizing and delineating building roof boundaries in image-space. In view of this, this project integrates several subprojects that have been recently started in my research group, i. e.: ALS data filtering; automated building boundary extraction from ALS data; automated building boundary extraction from very high-resolution aerial images; and automated building boundary extraction combining building roof boundaries previously extracted from both data sources. (AU)

Scientific publications
(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)
MARCATO FERNANDES, VANESSA JORDAO; DAL POZ, ALUIR PORFIRIO. Extraction of building roof contours from the integration of high-resolution aerial imagery and laser data using Markov random fields. INTERNATIONAL JOURNAL OF IMAGE AND DATA FUSION, v. 9, n. 4, p. 263-286, 2018. Web of Science Citations: 0.
MARCATO FERNANDES, VANESSA JORDAO; DAL POZ, ALUIR PORFIRIO. A Markov-Random-Field Approach for Extracting Straight-Line Segments of Roofs From High-Resolution Aerial Images. IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, v. 9, n. 12, 1, SI, p. 5493-5505, DEC 2016. Web of Science Citations: 2.

Please report errors in scientific publications list by writing to: cdi@fapesp.br.