Support Vector Machine Framework for Detecting Ambiguous Contours of Space Debris Cloud

Operational spacecraft are prepared for collisions with space debris by installing protective walls (bumpers). A group of vaporized and liquefied microparticles generated by debris collisions is called a debris cloud. To evaluate the protective performance of bumpers, it is important to measure the spread velocity and angle of the debris cloud. However, debris cloud images exhibit ambiguous contours of debris clouds. Consequently, measurement errors in the spread velocity and spread angles occur because of the observer’s subjectivity. Commonly used derivative-based methods can detect steep contours; however, they cannot detect ambiguous contours such as those of debris clouds. In this study, a method for uniquely detecting the ambiguous contours of a debris cloud based on supervised machine learning and a continuous wavelet transform (CWT) is proposed. Feature points that indicate candidate points of the debris cloud contour are extracted using a method based on CWT. Four types of specific features for debris cloud contours were assigned to the feature points. Supervised machine learning using a support vector machine can detect the debris cloud contours. The proposed method enables the evaluation of the ambiguous contours of a debris cloud and the protection performance of a bumper without observer subjectivity.