Co-registration of lunar imagery and digital elevation model constrained by both geometric and photometric information

• Published in: Planetary Remote Sensing and Mapping pp. 251 - 265
• Main Authors: Liu, B., Xin, X., Di, K., Jia, M., Oberst, J.
• Format: Book Chapter
• Language: English
• Published: CRC Press 2019
• Edition: 1
• Subjects: Harris Detector; Coarse DEM; Iterative Closest Point Algorithm; Solar Azimuth; Reference DEM; Simulated Image; Affine Transformation; LRO; High Resolution DEM; Image Matching; RFM; Image Id; Co-registration Process; PDS; Radiometric Information; Matching Precision; Image Matching Method; Proposed Matching Method; Low Resolution DEMs; Laser Altimeter; Tie Points; RSM; Rational Polynomial Coefficients; Solar Altitude; Matching Method
• ISBN: 9781138584150; 1138584150
• DOI: 10.1201/9780429505997-17

The alignment of images with a digital elevation model (DEM) has many applications in the planetary mapping field. In this chapter, we propose a novel highly precise co-registration method that can achieve direct pixel-based matching between an image and a reference DEM. The DEM is first converted into a simulated image using a hill-shading technique based upon a photometric model and the image’s illumination conditions. Initial matching between the simulated image and the input image is then performed based on affine scale-invariant feature transform (ASIFT). Meanwhile, the image’s rational function model is established and used as a geometric constraint. Next, the tie points generated by the initial ASIFT matching are used to refine the rational function model geometric model of the image and to eliminate the gross errors of tie points in an iterative fashion. Finally, highly precise co-registration is performed by pixel-based least-squares image matching using the refined geometric model as a global geometric constraint. Two Lunar Reconnaissance Orbiter (LRO) narrow-angle camera images located at the pre-selected landing site of the Chang’E-5 mission and the SLDEM2015, a combined product of lunar orbiter laser altimetry (LOLA) and DEM generated from the Japanese Selenological and Engineering Explorer (SELENE) terrain camera images, were selected in our experiment. The results demonstrate that the proposed method can achieve effective pixel-based matching between an image and a reference DEM with a mean accuracy of 15.6 pixels in the image space and 20.95 m (0.5 pixel of the reference DEM) in the object space. Another 15 narrow-angle camera images that were evenly distributed at various latitudes were selected as an additional experiment to evaluate the matching precision of various DEM densities. The results show that the precision improved in a near-linear manner as the latitude increased. Thus, the denser the DEM, the greater the matching precision. The proposed method offers a highly precise and automatic method of matching orbiter images with DEMs. This chapter presents a novel pixel-based method using both geometric and photometric constraints for co-registration of lunar orbital image and reference digital elevation models (DEM). High-precision co-registration of images and DEMs is needed to construct a higher-resolution DEM using a coarse DEM and a high-resolution image with shape-from-shading techniques. The rational function model (RFM) of the image is established based upon the rigorous sensor model and is used as a geometric constraint in the co-registration process. Seventeen simulated images are generated based upon the images’ illumination information. The precision of co-registration is represented by the fitting precision of affine transformation for RFM refinement. The precision of co-registration is represented by the fitting precision of affine transformation for RFM refinement. The precision of the co-registration can reach a sub-pixel level of the reference DEM, and the proposed co-registration method outperforms both the affine scale-invariant feature transform matching method and the traditional least-squares method.