Estimation of False Detections for Evaluation of Crater Detection

Introduction: A method for estimation of false detections for crater detection algorithms (CDAs) is proposed. In combination with known ground truth (GT) and other available analyses, the proposed method can improve evaluation of CDAs. Evaluation of CDAs: CDA applications include dating of a planetary surface [1], autonomous landing to planets [2] and asteroids [3], and advanced statistical analysis [4]. The most common evaluation of CDAs is visual comparison of detected craters with topography. More advanced evaluations include comparison with craters' catalogue(s) [5-7] and cumulative size-frequency distribution(s) [1] provided by planetary scientists, as well as the use of Receiver Operating Characteristics (ROC) [8-12]. For those evaluations, GT (locations and sizes of known craters as labeled by geologists) is required. Limitation of Known GT: For estimation of GT, a catalogue with 17582 craters was assembled, wherein each crater is aligned with MOLA topography and confirmed by three independent sources: (1) catalogue from N. G. Barlow et al. [13]; (2) catalogue from J. F. Rodionova et al. [14]; and (3) revised version of catalogue used in previous work [15]. The new GT