AUTOMATIC CITY MODELLING BY ROBUST LINE MATCHING 
USING HIGH VISION AIR-BORNE IMAGERY 
Y. Kunii, H. Chikatsu 
Tokyo Denki University, Department of Civil Engineering, 
Hatoyama, Saitama, 350-0394, Japan 
(kunii, chikatsu)@g-dendai.ac.jp 
Commission V, WG V/5 
KEY WORDS: Three-dimensional, Modelling, Matching, Sequences, Extraction, City, Aerial 
ABSTRACT: 
In order to perform city modelling using images processing procedures, line or feature extraction and stereo matching will be 
performed in general. However, there are some issues for efficient 3D object modelling. In particular, efficient line matching for 
reconstruction of objects such as building and so on are needed to be resolved. With this objective, this paper investigates a robust 
line matching method which include epipolar matching and object recognition. Consequently, automatic city modelling using high 
vision air-borne imagery became possible. 
1. INTRODUCTION 
Recently, efficient spatial data acquisition and visualization 
have been receiving more attention from the view point of 3D- 
GIS, city modelling and so on. Generally, in order to perform 
object modelling using digital images, line or feature extraction 
and stereo matching will be performed which have been 
proposed many methods such as area based matching, future 
based matching, and so on (Schenk, 2001). The authors have 
been concentrating on developing a 3D modelling method for 
buildings using video image sequences (Kunii, 2000). The 3D 
modelling method contains line matching which is performed 
by line extraction and line tracking. The line extraction was 
performed by Canny operation (Canny, 1986), and the line 
tracking was performed by optical flow estimation. However, 
for efficient city modelling by air-borne imagery, automated 
segmentation of feature areas or extraction of feature lines for 
man-made objects is still issue which is needed to be resolved. 
With this objective, a robust line matching method which 
include epipolar matching and object recognition were 
investigated in this paper. Consequently, more efficient line 
matching can be performed, and 3D data of the each both ends 
for the lines were calculated. Furthermore, recognition of 
  
objects became possible, and city modelling can be performed 
automatically. 
This paper reports the automatic city modelling method using 
HDTV imagery. 
2. HDTV IMAGERY 
HDTV (High Definition Television) is a new format of the 
television imagery which can be obtained by the special device. 
The HDTV imagery has 1125 scanning lines (NTSC: 525, PAL: 
625), and high resolution image sequences can be obtained. 
This paper reports the automatic modelling method using 
HDTV air-borne imagery which obtained from a helicopter at 
urban district of Kobe-city, Hyogo, Japan. Figure 1 shows the 
first frame in the HDTV imagery. 
3. LINE MATCHING 
Automatic Line matching was performed by line extraction and 
optical flow estimation. Detail procedures of the automatic line 
matching method are as follows. 
3.1 Line Extraction 
Line extraction was performed by Canny operation with 2 
threshold values relative to 2 components which called the 
height and reliability of edge (Canny, 1986). The height of edge 
is a variation of the gray level around at a interest point, and the 
reliability is an index for representing influence of noise. 
Furthermore, In order to perform line tracking, each both ends 
for the lines were connected by straight line. However, there are 
some useless lines for line tracking in the first image. Then, the 
useless lines were removed by manual operation in the only 
first image. Figure 2 shows the line extraction for the first 
image. 
—500—