Space debris has been continuously increasing with the advancement of space development in recent years. In particular, the measures against debris in the range of several hundred micrometers to several millimeters are inadequate. The main cause of this micro debris is secondary debris called as “ejecta” generated by space debris or meteoroid which collides with a spacecraft’s surface. Therefore, it is necessary to evaluate the size of ejecta from spacecraft materials, and impact experiments based on ISO 11227 have been conducted. This standard intends that the analysis of images from a witness plate placed near the target can be used to infer the size distribution of ejecta from the impact craters on the plate, serving as an evaluation criterion. This study aims to detect impact craters with high accuracy, regardless of their formation patterns or brightness, by using machine learning to automatically identify impact craters on a witness plate. Specifically, R-CNN model trained on various images of impact craters was used to identify them and edge detection using morphological image processing was performed to measure the size of craters.
Kawase et al. (Fri,) studied this question.