Three Dimensional Positioning for Airborne Three-Line-Scanner Images
Abstract
The purpose of this investigation is to perform 3-D positioning using three-line-scanner Level 1
images. Three major steps are included: (1) definition of the geometry of the triplet, (2)
calculation of orientation parameters, and (3) space intersection. In the first step, the exterior
orientation parameters are expressed as low-order polynomials with respect to time. The
additional parameters are used to represent image coordinates including forward and backward
images. Then, the orientation parameters are calculated by space resection. Finally, the results
are checked by space intersection. A set of ADS40 data covering the area of Waldkirch of
Switzerland is used for validation. The experimental results indicate that the proposed method is
simple and yet may reach high accuracy.
1. INTRODUCTION
Airborne three-line-scanner images have the merits of high spatial and spectral resolutions, and
excellent converging geometry. Thus, the images have become an important data source in
environmental remote sensing and GIS application. In order to extract 3D information from the
images, orientation modeling is a prerequisite. The bundle adjustment is often used to calculate
the dynamic orientation parameters with respect of time for three images (Shibasaki et al, 2003).
The method might contain local systematic errors for the data with high dynamics, thus, a least
squares filtering that performs orbit collocation is preferable.
Before calculation of bundle adjustment, the imaging geometry has to be constructed. At the
initial stage of this investigation, we perform the space resection to calculate the orientation
parameters. Then the orientation parameters may be used in three-dimensional positioning by
intersecting three rays for conjugate points.
In the space resection, we compare two different methods. The first one is the scene dependent
approach that means each image has its own orientation parameters. This simplified method
assumes that the three images have no geometric constraints. The second method considers that
the three images have the same orbit and attitude parameters. The CCD distances between
forward, nadir, and backward are pre-calculated, and we then use those values to connect three
images by including additional parameters.
The purpose of this investigation is to perform three-dimensional positioning using
three-line-scanner Level 1 images. Different from the raw data, the Level 1 images have been
preliminarily rectified using GPS and INS data. The distortions in the scenes of Level 0, caused
by the motion of the sensor, are mostly removed by this rectification (L. Hinsken, et al 2002).
Therefore, the images include small tilt displacements. The central tasks include: (1) definition
of the geometry of the triplet, (2) calculation of orientation parameters, and (3) calculation the
ground points by space intersection. The workflow of investigation is in figure1.
Figure 1. Workflow of the investigation
