Full text: Mapping without the sun

Jianming Gong a,b 1 *, Xiaomei Yang b , Chenghu Zhou b , Xiaoyu Sun b , Cunjin Xue b 
a School of Remote Sensing and Information Engineering, Wuhan University, Wuhan, Hubei 430079, P. R. China 
b State key Lab of Resources and Environment Information System, CAS, Beijing 100101, P. R. China 
Commission VI, WG VIM 
KEY WORDS: RFM, RPC, Beijing-1 Small Satellite, Orthorectification 
With two pushbroom sensors of which one is 4m panchromatic sensor, and the other is a 32m multispectral sensor, Beijing-1 small 
satellite have both the features of the high resolution of SPOT and the multi-spectrum of Landsat. High resolution satellite imageries 
(HRSIs) increase the need for higher accuracy of data modeling. The satellite orbit, position, attitude angles and interior orientation 
parameters have to be adjusted in the geometrical model to achieve optimal accuracy with the use of a minimum number of Ground 
Control Points (GCPs). However, most high resolution satellite vendors do not intend to publish their sensor models and ephemeris 
data. Therefore, a range of alternative and practical approaches should be developed for extracting accurate 2D and 3D terrain 
information from HRSI. Aimed at the features of Beijing-1 small satellite high resolution panchromatic imagery, an 
orthorectification method based on Rational Function Model (RFM) is proposed in the paper. RFM is a generalized sensor model to 
perform orthorectification in no need of orbit parameters and sensor imaging parameters which is independent on sensors or 
platforms and supports any object space coordinate system with a variable coordinate system. The improved RPM based on block 
adjustment with compensation for exterior orientation biases is then refined and evaluated, as is the means to enhance the original 
Rational Polynomial Coefficients (RPCs) through a bias correction procedure. Compared to linear transformation and polynomial 
transform, RFM has the highest positioning accuracy, because RPC is determined by applying the least squares principle to GCP 
data, approximate error can be evenly distributed through RFM rectification. An amount of experiments have proved that a sub-pixel 
positioning accuracy can be achieved by using refined RFM to rectify the Beijing-1 small satellite image which is close to the 
accuracy of the rigorous sensor model. 
In order to ensure well and truly utility for Beijing-1 HRSI data 
which has a high resolution of 4m, a higher rectification 
accuracy of image processing has been taken into account 
(Chen Zhengchao, 2006). Besides geometric accurate correction, 
sensor model must be established to meet the needs of high 
accuracy in orthorectification. There are two categories of 
sensor models: physical sensor models and generalized sensor 
models (C.S. Fraser, 2000; Okamoto Atsushi, 1999; Toutin T., 
2002). Establishment of physical sensor models requires the 
information such as the physical structure of the sensor and the 
imaging mode. However, for secrecy, information such as the 
lens structure, imaging mode and orbit parameters of some high 
performance sensor has not been made public. Therefore, the 
users usually can not build the rigorous models of these sensors 
and require generalized sensor models that are independent on 
sensors with a simple form and quick processing. RFM, linear 
equations, and polynomial transform are the typical examples 
of the generalized sensor models. Many kinds of senor models 
and their accuracy have been studied by researchers at home 
and abroad. Yong Hu et al from GeoICT, NYU and C.V. Tao 
from Canada have done important research (Yong Hu, 2004; 
C.V. Tao, 2001a; C.V. Tao, 2001b). Shulong Zhu from Hong 
Kong Polytechnic University (Zhu shulong, 2004), Zhian 
Zhang from Taiwan Central University (Tee-Ann Teo, 2005), 
Xuwen Qin, Guo Zhang from Wuhan University (Qin Xuwen, 
2005), and Yongsheng Zhang from Information Engineering 
University all have done research on generalized sensor models 
and accuracy evaluation (Zhang Yongsheng, 2004; Liu Jun, 
2002). Research achievements available indicate: the accuracy 
of RFM is better than that of polynomial model, and in regions 
with small change of the terrain, RFM can use fewer GCPs to 
achieve ideal accuracy. The sensors of the Beijing-1 high 
resolution satellite have the key features of a long focal length 
and a narrow field of view. Especially, under the condition of 
the focal length of the panchromatic sensor is 1.371m, and the 
field of view is 1.9 degrees RFM becomes prominent. Hence, 
the RPC generalized sensor model is proper for the Beijing-1 
high resolution satellite. Since RFM has considered the 
influence of the height change ignored by traditional 2D 
transformations, it is more accurate than the polynomial model 
and spline function model (C.S. Fraser, 2006). Meanwhile, 
without knowing the information about the sensor imaging 
process, users can use simple RFM to rectify the relation 
between image space and object space accurately. In the 
development of Beijing-1 small satellite data processing 
software, using a refined RFM to perform orthorectification on 
the Beijing-1 small satellite data has achieved satisfactory 
results (J.M. Gong, 2007). 
* Corresponding author: Jiangming Gong, E-mail: gongjm@lreis.ac.cn. This work was supported by a grant from the Major State 
Basic Research Development Program of China (973 Program) (No.2006CB701305) and the National High Technology 
Research and Development Program of China (863 Program) (2006AA12Z146; No. 2005AA133013 ) .

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.