A Review of Impact-Crater Detection

Currently many methods and algorithms have been applied for the impact-crater detection, and the achieved accuracies and adaptabilities are not quite similar among different approaches and data. This paper first summarizes the current progress of the area, and then discusses the advantages/disadvantages of different approaches and their applicable conditions. We finally analyze the main problems in the research of impact-crater detection by pointing out possible solutions. Current crater-detection approaches fall into four categories: 1) Manual recognition, 2) Shape-profile fitting algorithms including Hough-Transformation, conic-fitting, and template-matching algorithms, 3) Machine-learning methods including SVM, genetic algorithm, and neural network methods, and 4) Geological-information based analysis using terrain data and spectral data. From the comparison between different approaches, we derive the following conclusions. 1) Manual recognition is suitable for cases with only image data available. The accuracy depends on the experience of researchers, and the efficiency is low. 2) The shape-profile fitting algorithms are suitable for the craters with simple structures and clear edges. 3) The Geological-information based analysis does not depend on image quality, but instead depends on illumination and resolution. With more and more highly accurate data, the crater detection will tend to use several types of data and combine results from these.