Presentation #211.05 in the session We Know the Way: Future Missions, Instruments, Facilities (iPosters).
A prototype of a novel dual sensor multispectral imager has been developed and tested by our group at the University of Western Ontario as part of a Canadian Space Agency (CSA)-funded project for lunar rover-based applications. The instrument features a visible to near-infrared (VIS-NIR) CMOS sensor combined with a short-wave infrared (SWIR) InGaAs sensor. It has been designed to capture images in several spectral bands from 400 nm to 1700 nm, providing the capability to characterize igneous minerals relevant to lunar volcanic and cratering processes, as well as the presence of water ice.
The use of two different types of sensors provides unique challenges in accurately aligning the various spectral images obtained from each sensor. The varying optical characteristics across the broad wavelength range in this instrument make post-processing necessary to co-register the images accurately, without which artifacts such as color fringing are observed. This co-registration must account for the different sensor geometries, pixel sizes, focal lengths, and lens distortion effects between the VIS-NIR and SWIR sensors, which are also physically separated by a non-trivial offset. A co-registration pipeline is necessary to ensure the spectral integrity of the resulting data cube.
We have adapted co-registration algorithms typically used in medical imaging to develop an automated co-registration technique to align the images obtained from each spectral band. The technique was tested on multispectral images of anorthosite and basalt lunar analog samples obtained under natural solar illumination. We used both visual evaluation of color fringing and image similarity metrics to evaluate co-registration quality. We are currently studying the effects of correcting for optical distortion using the camera calibration parameters on the quality of the final co-registered spectral cube.
The results show good alignment between images taken with the same sensor, with little to no color fringing. The results also show that good alignment is obtained between images taken by the VIS-NIR and SWIR sensors, but only over the central part of the scene occupied by the analog samples.
The authors would like to acknowledge funding from the Canadian Space Agency (CSA) and NSERC CREATE (SMART-ART.)