Integration of Light Curve Brightness Information and Layered Discriminative Constrained Energy Minimization for Automatic Binary Asteroid Detection

Binary asteroid exploration is an important research focus in the areas of deep space exploration and space informatics because of its unique scientific value in analyzing the structure and gravitational dynamic properties of asteroids, as well as the origin and evolution of celestial bodies in the solar system. Remote photometric observations can reveal the key characteristics of binary asteroids that differ from other kinds of asteroids, especially unary asteroids, and provide an efficient and convenient way to discover binary asteroids. However, the automatic binary asteroid detection problem by advanced machine learning methodology remains unresolved when handing complex asteroid photometric data. For this problem, this article proposes to simulate unary and binary asteroid systems using cellinoid and oblate sphere shape models and generate the corresponding light curve brightness information with different asteroid physical parameters. Then, a benchmark unary and binary asteroid light curve dataset is constructed. Afterward, a layered discriminative constrained energy minimization (LDCEM) method is developed to train a binary asteroid detector in a layered learning manner by enlarging the response differences between the light curve data of unary and binary asteroids such that the two common kinds of asteroids can be well distinguished. The experimental results on the simulated asteroid light curve data, as well as the real observed asteroid light curve data, show that the proposed LDCEM method can yield promising binary asteroid detection performance in comparison with some representative detection methods.