THE GEOMETRIC ACCURACY VALIDATION OF THE ZY-3 MAPPING SATELLITE

ZiYuan-3 (ZY-3) mapping satellite is the first civilian high-resolution stereo mapping satellite of China. The satellite’s objective is oriented towards plotting 1:50,000 and 1:25,000 topographic maps. This article proposes ZY-3 mapping satellite Rigorous Image Geometry Model and Rational Function Model (RFM). In addition, this paper utilizes the image of the ZY-3 satellite with the region of flatlands, hills and mountains for the block adjustment experiment. Different ground control points are selected and the accuracy is validated by check points, and the some Digital Surface Model (DSM), Digital Orthophoto Map (DOM) are generated and the accuracy is also validated by check points. The experiment reveals that the planar accuracy of DOM and vertical accuracy of DSM are better than 3m and 2 m, respectively. The experiment demonstrates the effectiveness of ZY-3 mapping satellite image geometry model. * Corresponding author. This is useful to know for communication with the appropriate person in cases with more than one author.


INTRODUCTION
The ZiYuan-3 (ZY-3) mapping satellite, launched on January 9, 2012 in Taiyuan Satellite Launching Center, is the first civilian high-resolution stereo mapping satellite of China.Its objective is oriented to plot 1:50,000 and 1:25,000 topographic maps.It is an optical satellite with Triple Linear-array Camera and Multispectral Camera in push-broom imaging mode.
Triple Linear-array Camera is constituted with three highresolution panchromatic cameras working in timedelay integration mode (TDI CCD).The panchromatic cameras are, respectively, in the nadir view, forward view and backward view.The forward view and backward view cameras both have a spatial resolution of 3.5m and 52km ground swath while the nadir view camera has a spatial resolution of 2.1m and 51km ground swath.Multispectral Camera provides multi-spectral 5.8m resolution images with 51km ground swath.ZY-3 offers more than one year service of space teledetection to a wide variety of users.The data have been widely applied in surveying and mapping, agriculture, forestry, environmental protection, disaster reduction, urban planning and other departments at present.And the application department also increase rapidly.Up to now, It has been distributed more than 35 million square kilometers to domestic institutions, academy, university and other country such as American, German and Australia.

ZY-3 SPECIFICATIONS
ZY-3 (see Figure 1) is being manufactured by China Academy of Space Technology and its primary user is National Administration of Surveying, Mapping and Geoinformation.Table 3.The Specifications of Multispectral Camera

THE GEOMETRY MODEL OF ZY-3 MAPPING SATELLITE
The geometry model of images indicates the mathematical relationship between image coordinates (x, y) and ground coordinates (X,Y,Z) (Toutin 2004, Zhang 2005, Zhen and Zhang 2005, Zhu et al. 2009), involving two types.One is the rigorous image geometry model and the other is the general image geometry model, also called rational function model(RFM).

Rigorous Image Geometry Model
We constructed the rigorous image geometry model of ZY-3 calibrated images based on the gazing direction vector at the imaging moment of CCD linear array.We then established oneto-one correspondence between image point and ground point by calculating the intersection point between the CCD detector gazing direction and the earth ellipsoid model (Riazanoff 2004, Zhang 2005, Zhu et al. 2009).
As far as the raw data of ZY-3 are concerned, GPS measures the position of GPS phase centre, and attitude sensor measures the orientation from star sensor to ground in the J2000 inertial frame.In order to acquire the camera position and attitude of the main optical axis, the data measured by GPS and star sensor need to be transformed to the position and orientation of the camera.Therefore, it is necessary to measure three offset vectors of the GPS phase centre in the satellite's body-fixed coordinates ,the rotation matrix

Rational Function Model
To facilitate the data processing of ZY-3, both the rigorous image geometry model and the RFM are constructed in this study.
Rational Function Model use the ratio polynomial to establish the relationship between image coordinates (x, y) and ground coordinates (X,Y,Z).In order to enhance the stability of the parameters calculation, the image coordinates and ground coordinates will be normalized to [-1,1], the ratio polynomial is established as follows: (2) According to the steps solving the RFM parameters (OGC 1999, Tao and Hu 2001, Zhang 2005), the RFM parameters of ZY-3 image are calculated as follows: 1.According to the image coverage, calculate the maximum and minimum ellipsoid elevation of sensor corrected product using globe 1km resolution Digital Elevation Model(DEM). 2.Then, establish ground regular grid in a certain size and generate ground coordinates of control points.
3. Calculate image coordinates of control points using anti-transformation model of rigorous geometry model of sensor corrected product.4. Calculate the Rational Function Model by least squares adjustment and then validate using the check points.

Geometric Accuracy Analysis
The paper utilizes the imagery of the ZY-3 satellite with the region of flatlands, hills and mountains for the experiment.Different ground control points are selected and the accuracy is validated by check points.

Uncontrolled Block adjustment:
Free block adjustment, all of control points can be seen as check points to evaluate the accuracy.Using a few control points do block adjustment, others can be seen as check points to evaluate the accuracy.4 and 5 that the mean square error of planar was better than 10m and the mean square error of height was better than 8m without control points.With a few control points, the mean square error of planar was better than 3m and the mean square error of height was better than 2m.

DSM and DOM product
After block adjusting the forward, backward and nadir imagery with four GCPs, the DSM was produced using dense matching techniques (Jiang 2004, Zhang and Zhang 2008, Zhang et al. 2011b) (Figure 8).An ortho-rectified image was consequently produced (Zhang 2005) (Figure 9).The remaining 470 GPS points (Anping) were used as check points to evaluate production accuracy.Vertical root mean square (RMS) error of the DSM was 2.01 m, while the planar RMS error of the orthoimage was 1.34 m.

CONCLUSION
It can be seen from the experiment that the planar and vertical accuracy are better than 3 meters and 2 meters respectively with a few control points, matching the criteria of 1:50,000 topographic mapping.The experiment results demonstrate the effectiveness of the ZY-3 surveying satellite imaging geometry model.
In light of the adjustment results and accuracy assessment of the DSM/DOM, ZY-3 linear-array images have reached acceptable planar and vertical geometric accuracy compared with the images of mature foreign commercial satellites of the same resolution.ZY-3, which has the broad application prospects, sets off a new era of transition from qualitative geometry to quantitative high-precision geometry for domestic satellites.Now, ZY-3 data have been widely applied in surveying and mapping, agriculture, forestry, environmental protection, disaster reduction, urban planning and other departments at present.It provides stable data source to plot 1:50,000 surveying and mapping products, update 1:25,000 even larger scale maps, and carry out mapping and change detection of land resources.In addition, the image data provide emergency mapping services to help disaster management, ecological protection, national geographical state monitoring, etc.

Figure
Figure 1.flight status of ZY-3 frame and the satellite platform, as well as the pointing angle (X，Y) of every CCD array detector.Regardless of the atmospheric refraction effects, the rigorous image geometry model for ZY-3 calibrated image is established as follows (Xingming Tang 2013): -dimensional Cartesian coordinate of the ground point P in the WGS84 frame, m is the proportional coefficient, and f is the focal length of the camera lens.

Figure 2 .
Figure 2.Control point distribution of Xianning

Figure 5 .
Figure 5.Control point distribution of Taihang Mountain

Figure
Figure 8.DSM Product

Table 5 .
Controlled Block Adjustment Result It can see from Tables