ASSESSMENT OF FEATURE DETECTORS AND DESCRIPTORS IN REMOTE IMAGES OF PLANETARY BODIES
Algorithms to detect, describe, and match common features in image sets are expanding to new worlds with the integration of the OpenCV feature matching framework in a popular mapping tool for planetary images (ISIS3). These algorithms provide a new approach to register images and build image-based control networks. However, the natural landscape of the Moon and other planetary bodies pose new challenges such as numerous features that appear similar (i.e., impact craters and boulders). In addition, planetary image sets vary in scale, orientation, and noise properties, especially when conducting cross instrument and cross mission comparisons. This study assesses a collection of common feature detector and descriptor algorithms to examine how they adapt to these challenges. With our analysis, we did not identify an ideal detector and descriptor combination that exists for our diverse lunar dataset. However, we did identify where particular algorithms succeed and identify their shortcomings. By knowing these capabilities, users can identify the proper set of algorithms to apply to an image set given the presence of noise and variations to scale and orientation.