The high resolution lunar global images acquired by Chang'E-2 (CE-2) CCD camera orbiter are of great importance for
lunar science research as well as preparation of the landing and surface operation of Chang'E-3 (CE-3) lunar rover. In
this paper, a rigorous geometric model of CE-2 CCD camera is developed based on the push-broom imaging principle. A
self-calibration bundle adjustment (SCBA) method is proposed to eliminate the inconsistencies between forward- and
backward-looking images, which are caused by the inaccuracy of exterior orientation (EO) parameters and the uncertain
relationship between the two CCD line arrays of the camera. The interior orientation (IO) model is refined by adding
several additional parameters and the EO parameters are fitted by a third-order polynomial model. Strategies for ensuring
the robustness and reliability of the solution are also adopted, including EO pseudo observations selection, reasonable
weight determination, and truncated singular value decomposition method. After adjustment, the inconsistencies between
the forward- and backward-looking images are eliminated effectively by reducing the image-space residuals from around
20 pixels to sub-pixel. Based on the adjustment, high precision DEM (Digital Elevation Model) and DOM (Digital Ortho
Map) of a local area at Sinus Iridum (pre-selected CE-3 landing site) are generated automatically.
Chang'E-1(CE-1) is the first lunar orbiter of China's lunar exploration program. The CCD camera carried by CE-1 has acquired stereo images covering the entire lunar surface. Block adjustment and 3D mapping using CE-1 images are of great importance for morphological and other scientific research of the Moon. Traditional block adjustment based on rigorous sensor model is complicated due to a large number of parameters and possible correlations among them. To tackle this problem, this paper presents a block adjustment method using Rational Function Model (RFM). The RFM parameters are generated based on rigorous sensor model using virtual grid of control points. Afterwards, the RFM based block adjustment solves the refinement parameters through a least squares solution. Experimental results using CE-1 images located in Sinus Irdium show that the RFM can fit the rigorous sensor model with a high precision of 1% pixel level. Through the RFM-based block adjustment, the back-projection residuals in image space can be reduced from around 1.5 pixels to sub-pixel., indicating that RFM can replace rigorous sensor model for geometric processing of lunar images.