DocumentCode
2981207
Title
Study on outline extraction for ground buildings in high resolution SAR images
Author
Li, Yan ; Cao, Xiaoguang ; You, Hongjian
Author_Institution
Image Process. Center, Beihang Univ., Beijing, China
fYear
2009
fDate
26-30 Oct. 2009
Firstpage
813
Lastpage
816
Abstract
This paper analyses the visual characteristics of ground buildings in high resolution SAR images: the L-shaped echoes resulted from layer over and corner reflection effects give expression to the information of location and height of buildings. Therefore, to vectorize the L-shaped outline of the buildings has a great value in the field of city mapping. Focusing on L-shaped echoes, this paper investigates an outline vectorization algorithm for buildings. Firstly, the method counters the image after segmentation, then gets the main direction of buildings based on Hough Transform. After using polygon approaches on the outline, finally, the method adopts piecewise linear fitting by least-square algorithm and constructs 3-D model of buildings. Experiments on SAR images show that this approach is much fast in extraction of ground buildings with convenience and practicality.
Keywords
Hough transforms; image segmentation; least squares approximations; radar clutter; radar imaging; synthetic aperture radar; Hough Transform; L-shaped echoes; corner reflection effects; ground buildings; high resolution SAR images; image segmentation; layer over reflection effects; least-square algorithm; outline vectorization algorithm; piecewise linear fitting; Buildings; Cities and towns; Counting circuits; Data mining; Image analysis; Image resolution; Image segmentation; Information analysis; Piecewise linear techniques; Reflection; Hough Transform; L-shaped echoes; Polygon approach; Vectorization;
fLanguage
English
Publisher
ieee
Conference_Titel
Synthetic Aperture Radar, 2009. APSAR 2009. 2nd Asian-Pacific Conference on
Conference_Location
Xian, Shanxi
Print_ISBN
978-1-4244-2731-4
Electronic_ISBN
978-1-4244-2732-1
Type
conf
DOI
10.1109/APSAR.2009.5374180
Filename
5374180
Link To Document