A study of management system of the ground control point database for ROCSAT-2 Remote Sensing images
Chih-Li Chang
National Space Program Office
8F, 9 Prosperity 1st Road, Science-based Industrial Park
Hsinchu, 300, China Taipei
Tel: +886-3-5784208 ext. 8461
Fax: +886-3-5799972
E-mail: CLChang@nspo.gov.tw
Ta-Ko Chen
Researcher , Remote Sensing Dept.
Energy & Resource Laboratories ,ITRI
Bldg. 24, 195-6 Sec. 4, Chung Hsing Rd.
Chutung Hsinchu, China Taipei 310
Tel:+886-3-5914332
Fax:+886-3-5820038
E-mail: Tkchen@itri.org.tw
Abstract
To precisely correct geometric distortion of ROCSAT - 2 remote sensing images, a database
of ground control points (GCPs) is urgently needed for China Taipei area.
A known algorithm of geometric correction was introduced first. And a distribution of GCPs of China Taipei
area was proposed. The distribution of GCPs is to make sure that any scene of China Taipei will be
allocated enough GCPs for the data processing. Then, the attributes of a GCP are defined and the
functions of a management database system are described, accordingly.
Finally, three GCPs were selected as test samples. They are located in the north, middle and south of
China Taipei, respectively. They represent for the city, country and mountain areas, respectively. The
attributes of the three selected GCPs were collected on ground. The north testing GCP in the database
management system was demonstrated in the study.
Introduction
The ROCSAT- 2 program was initiated in 1997. Mission of the ROCSAT -2 program was defined as below.
“ROCSAT - 2 mission objectives include civil applications to natural disaster evaluation,
agriculture and forestry, land use, environmental monitoring, coast search and rescue,
academic researches and education, and international cooperation.” (Wang, 2000)
An investigation on the requirements and applications of remote sensing was conducted at the beginning
(Chang 2002a). Two kinds of studies for ROCSAT-2 program have been carried out. First, the image
processing technique has been studied. To estimate rice production, canopy reflectance spectrum was
measured in the test field. The yield data were recorded. The correlation between yield and spectral
reflectance was studied. The DEM can be derived from a stereo pair of simulated ROCSAT -2 images.
The ROCSAT - 2 images were simulated from higher resolution sources. The sources are from air and
space images. The identification of rice field has been studied extensively in China Taipei. In practice,
the raw data with cloud cover less than 10% is required to be processed into product for an optical remote
sensing image. More frequently visiting the same target will be more chance to get good quality of images.
Since ROCSAT-2 is the only daily revisit remote sensing satellite so that it will provide the most images
among optical imaging satellites for the agriculture applications in China Taipei.
Second, ground truths data were collected. Spectral reflectance of the mainly land cover/use was.measured with hand- held spectrometer. (Chang 2002b) The simulated ROCSAT -2 spectral reflectance
was derived from the ground measurement. At the same time, the locations of ground control points were
surveyed for precisely geometrical correction in China Taipei.
Ground Control Points
To correct radiometric and geometrical distortion for a level-1b product, the prior known parameters are
applied for data processing. The ROCSAT- 2 satellite has the very good pointing capability with the stable
attitude determination and control system. However, The GCP is needed as the ground reference to
precisely correct the location accuracy of the image.
A location model was built for level- 1b processing (CNES 1997). The location model is a pair of quadratic
equations to present the relation between the number of pixel and the coordination of the target point.
Two models such as direct and inverse models are specified with twelve coefficients.
where:
x and y are the line and column number of the pixel
and are the latitude and longitude of the target point
To reduce the number of the coefficients, a simplified model is specified with fewer coefficients, for
example, eight coefficients. For eight-coefficient model, the coefficients of e, f, e’ and f’ are neglected.
If the models are applied to precisely geometrical correction with GCPs, four to six points are required to
calculate the coefficients of the models. To design the distribution of GCPs for China Taipei area, some
assumptions of distribution for the GCPs are as follows.
- A typical ROCSAT-2 image scene with size of 24 km x 24 km
- At least four GCPs in a ROCSAT-2 image scene
- Start line of a scene ahead of a grid line for 1 km
- GCP around the corner of the grid within a radius of 1 km
One of the grid systems was studied with the size of 11km × 11km in China Taipei. Then, any typical
scene will cover six to nine points of GCPs so that more than minimal number of four can be applied for a
precisely geometrical correction.
