Aluir Dal Poz
AUTOMATIC SPACE RESECTION USING A CONSTRAINED RELATIONAL MATCHING
Aluir Porfírio Dal Poz'
Antonio Maria Garcia Tommaselli”
"Sao Paulo State University - Pres. Prudente, SP - Brazil
Department of Cartography
{ aluir, tomaseli } @prudente.unesp.br
Working Group III/1
.KEY WORDS: Digital Photogrammetry, Straight Lines, Automation, Relational Matching, IEKF.
ABSTRACT
Image orientation is a basic problem in Digital Photogrammetry. While interior and relative orientations have been
successfully automated, the absolute orientation (or space resection) continues to be an important topic for research. An
approach has been developed to automate the absolute orientation based on relational matching and a heuristic that uses the
analytical relation between image and object-space straight lines. A build-in self-diagnosis is also integrated in this
method, involving the implementation of data snooping statistic test in the process of spatial resection using the Iterated
Extended Kalman Filtering (IEKF). The aim of this paper is to present the basic principles of the proposed approach and
results based on real data.
1 INTRODUCTION
Considerable progresses have been accomplished in the automation of several geometric tasks, as, for example, the interior
and relative orientation and the generation of digital orthophoto. However, the situation is quite different for semantic
tasks, like the linear feature extraction from digital images for GIS data capture and updating and absolute orientation.
Concerning this last one, the main difficult is related to the fact that a correspondence process needs to be performed
between a digital image and a symbolic model describing the ground control. In such a case, line-based methods are
potentially better, because lines are easier to be detected in digital images and grouped than points.
In this paper, we are interested in a special class of lines, i.e., the straight lines. A simpler photogrammetric model can be
derived for this type of entity, whose complexity is similar to the well-known collinearity equation. Some straight line-
based photogrammetric model have been developed and can be found, for example, in Tommaselli and Lugnani (1988),
Mulawa and Mikhail (1988), Tommaselli and Tozzi (1996), and Tommaselli and Dal Poz (1999).
Although linear features are easier to locate than points and can be determined with sub-pixel precision, the automatic
feature extraction and correspondence is a difficult task to solve properly. A combination of several approaches is proposed
in this paper. Image orientation is recursively improved using IEKF (Iterative Extended Kalman Filtering) and the feature
extraction process is constrained by the filter feedback. This process was firstly applied in Machine Vision (Tommaselli
and Tozzi, 1996) but the feature extraction and the matching methods were not suitable to aerial images. Since then, a
relational matching method has been developed (Dal Poz et al, 1996; Dal Poz and Tommaselli, 1998) and feature
extraction algorithm has been improved as well (Tommaselli and Dal Poz, 1999).
The proposed solution is described in the section 2. The results based on real data are presented and discussed in the
section 3. Finally, conclusions are given in the section 4.
2 THE PROPOSED SOLUTION
The basic input data are the digital image, the ground control groupings (A, ..., A,), and the interior and approximate
exterior orientation parameters. Ground control groupings correspond to local structures (e.g., road crossing) and can be
extracted by measuring two 3D endpoints by conventional field survey or by photogrammetric plotting. The automatic
orientation process is carried out in three main steps. First, a ground control grouping is selected (e. g., A;) and its
position in the image-space is predicted and the feature extraction process is applied only to a small window enclosing
206 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.