Test Of ZY-3 Image Object Positioning Based On Rational Function Model

The Ziyuan-3 (ZY-3) satellite is China's first civilian high-resolution stereo mapping satellite, which is equipped with three-line-cameras (TLCs), looking forward, nadir and backward, respectively. The TLCs are used to acquire images covering the same area from three different observation angles. The ground sample distance (GSD) of the nadir camera is 2.1 m while the GSDs of the forward and backward cameras are both 3.5 m. With the purpose of improving the accuracy and the reliability of the direct object positioning of ZY-3 TLC images, two dual-frequency global positioning system (GPS) receivers, three star trackers and a couple of gyroscopes are installed on the ZY-3 satellite to measure ZY-3's position and attitude information. Nevertheless, affected by a number of factors, such as installation process and excessive acceleration during the satellite's launch, the values of the laboratory-calibrated orientation parameters, such as the positions of the cameras and the star trackers in the satellite-body coordinate system and the position of the charge-coupled device (CCD) detectors on the focal plane, often differ more or less from their in-orbit true values. By using the laboratory-calibrated orientation parameters, the direct object positioning accuracy of ZY-3 TLC images is about ±1.5 km in planimetry and ±270m in height. Therefore, a series of in-orbit geometric calibration approaches for ZY-3 TLCs were proposed in order to improve the direct object positioning accuracy of ZY-3 TLC images. As a result, by using the in-orbit geometric calibration, the direct georeferencing accuracy of ZY-3 TLC images was improved to be better than ±15m without ground control points (GCPs). Nevertheless, such an accuracy is still unsatisfactory for the high-precision needs for the surveying applications of ZY-3 TLC images. Therefore, it is necessary to make further improvement on the object positioning accuracy of ZY-3 TLC images.Taking the rational function model (RFM) as the geometric imaging model of ZY-3 TLCs, this paper focuses on the accuracy verification test of the object positioning of ZY-3 TLC images. Some main contents are shown as follows.1) Test of precise solution of the rational polynomial coefficients (RPCs). A large amount of virtual control points (VCPs) are firstly generated according to the physical sensor model (PSM) of ZY-3 TLCs. Then, the error equations are established by using the VCPs mentioned above. Finally, the ridge estimation is employed to solve the ill condition of the normal equations, so that the RPCs can be solved precisely. From the experimental results of ZY-3 TLC images covering the area in Luoyang, China, it can be seen that using the RFM to fit the PSM of ZY-3 TLCs can achieve an accuracy of better than 0.6 pixel.2) The test of compensating the systematic errors in the object positioning of ZY-3 single-view image. Based on the analysis of the systematic errors of the RPCs, a systematic error compensation model for the object positioning of ZY-3 single-view image is established by using the alllne transformation model to compensate the systematic errors of the RPCs, and the precise object positioning of ZY-3 single-view image is achieved. From the experimental results of ZY-3 single-view image covering Taiyuan, Luoyang, Xianning and Nanyang re-spectively, it can be seen that the affine transformation model can eliminate the effects caused by the systematic errors of the RPC on the object positioning of ZY-3 sing-view image, and the object positioning accuracy is improved to be better than 1.5 pixel.3) The test of the block adjustment of ZY-3 multi-view images. ZY-3 TLCs can acquire triple-overlapped images covering the same area in almost the same time. Hence, this study establishes a block adjustment model for ZY-3's multi-view images by using the inherent restrains of the triple-overlapped images.The experimental results of the block adjustment accuracy covering Taiyuan, Luoyang, Xianning and Nanyang images respectively, can achieve 2.7m,3.5m,3.6m,2.0m in planimetry and 2.6m,1.7m,2.0m,1.7m in height, which can fulfill the requirements of mapping in scale of 1:50000.