ASSESSMENT OF A DEVELOPED COMBINED POINT/LINE-BASED
PROJECTIVE EQUATIONS
H. F. Barakat“, H. A. Emam 5 M. S. Abdel- Wahab "
* Public Works Department, Faculty of Engineering, Cairo University, Giza, Egypt - hhbarakat@hotmail.com
? Survey Research Institute, National Water Research Center, Talbia, El Ahram Street, Giza, Egypt —
hany a emam(ghotmail.com, engmohamed25(ghotmail.com
Commission III, WG IIU/2
KEY WORDS: Mapping, Development, Comparison, Rectification, Satellite Imagery, Points, Straight Lines
ABSTRACT:
This paper presents the results of the assessment of the developed techniques for line-based and combined point/line-based eight
parameters projective equations. The developed equations were implemented in a general least squares adjustment technique where
both lines and/or points can be used for the rectification of both aerial and satellite images.
Several experiments were performed using the developed line-based and combined point/line-based techniques to rectify the various
types of satellite images and also aerial photographs. The experiments always started first with rectification of aerial photographs
before applying to satellite imagery, since the developed projective equations were derived based on frame geometry assumptions.
Then, the experiments were performed using several types of satellite imagery, such as Landsat7, SPOT4, IRS-1D, and IKONOS. In
all experiments, comparisons were performed between line-based and the commonly used point-based projective equations. It was
found that mostly similar results were obtained from both techniques. However, with points, usually a large number of control points
are needed to achieve accurate results (around 1-2 pixel resolution). Approximately, around 30 control points are needed for most
cases of satellite images. On the contrary, with lines, only 5 control lines are needed to achieve the same accuracy figures of 1-2
pixels. This can be justified since control lines can be considered as sequence of control points, and thus a line will provide much
more control information than a point.
The above conclusion presented a great reduction in the number of control features required for satellite image rectification. It
provides a great save in the field collection process and, thus, a more economic technique for mapping from satellite imagery. In
addition, the combined point/line-based developed technique will provide the flexibility, and thus speed, needed for the satellite
images rectification process through the use of any available control features. Finally, conclusions and recommendations for future
research were summarized.
1. INTRODUCTION completeness of the paper. In addition, least square adjustment
programs were developed using Mathlab version 6 for the
The use of satellite imagery for mapping applications has seen point-based, line-based, and combined point/line-based
great interest in recent years, especially with the availability of projective equations.
high-resolution satellite images of IKONOS and QuickBird.
Therefore, the need for using fast, economic, and, obviously, The well-established point-based eight parameters projective
accurate mathematical models for satellite images rectification equations have been commonly used for image rectification
has also increased. process. In this paper, the newly developed techniques of line-
based and combined point/line-based projective equations are
Generally, photogrammetic methodology, and in particular image used for the image rectification as well. Comparative analysis
rectification, has been based primarily on point features. of the results of the three techniques was also discussed.
Recently, line-based techniques have seen increased interest. This
is because straight lines are usually easier than points to extract In section3, results from several experiment using aerial
from digital imagery using automatic algorithms. Also, the use of ^ photograph and satellite images are provided for all three
straight lines makes it possible to use different line segments on techniques of point-based, line-based, and combined
the various overlapping images and consequently, eliminates the point/line-based projective equations. Various sources of low,
need for exact correspondence as in the case of points. In medium, and high resolution satellite images (Landsat7,
addition, straight lines are generally more available than points in SPOT4, IRS-ID, and IKONOS) were used to test the
imagery of human infrastructure. The inclusion of linear features, ^ performance of the developed techniques, especially
alone or in a combination with point features, into concerning the effect of using straight lines for the
photogrammetic reduction algorithms requires careful rectification of those time-dependant sensor imagery. Ground
development and analysis. control features (points and/or lines) were obtained using
either Global Positioning System (GPS) measurements or
In a previous work by one of the authors (Barakat 1997; Barakat ^ Egyptian Survey Authority maps of scale 1:50,000 and
et al 1995), new development of line-based projective equations 1:25,000.
and combined point/line-based projective equations were
introduced. Mathematical formulations of the developed Finally, conclusion is drawn regarding the performance of the
equations are presented in section 2 in this paper, along with the developed techniques, and recommendations made for future
classical point-based projective equations, for the sake of research.
184
International A
2.1 Point-Base
In the case of |
projective equa
in one plane tc
plane. Using ps
can be construc
ext fiv
ex fy
_ 82 + fay+
ext fy
Where X.
X. y:
(e,..J
The eight parai
points with kno
three of which
unknown coordi
known coordin
equations and th
The equations d
transformation.
was developed i
(e. f. E ,25)
point.
2.2 Line-Based
A very importar
of duality. In the
dual. Any theor
vice versa. By
may be used t
equation using |
form,
gx +
Where a2, b.
1
Equations (1) a
point correspond
for two protectiv
' ]
a =-
]
' r
EM
This equation is
3 equations (two
(3) is obtained us
