Asteroid shape inversion with light curves using deep learning

Context. Asteroid shape inversion using photometric data has been a key area of study in planetary science and astronomical research. Specifically, researchers have focused on developing techniques to reconstruct 3D asteroid shapes from light curves. This process is crucial for gaining deeper insights into the formation and evolution of asteroids, as well as for planning human space missions. However, the current methods for asteroid shape inversion require extensive iterative calculations, making the process time-consuming and prone to becoming stuck in local optima. For missions that aim to make a close approach to an asteroid, a faster and more efficient method is urgently needed. Aims. The goals of this work are to improve the precision, speed, and adaptability to sparse data in asteroid shape inversion and to support autonomous decision-making for shape inversion in space missions. Methods. We directly established a mapping between photometric data and shape distribution through deep neural networks. In addition, we used 3D point clouds to represent asteroid shapes and utilized the deviation between the light curves of non-convex asteroids and their convex hulls to predict the concave areas of non-convex asteroids. Results. With our approach, we eliminate the need for extensive iterative calculations, achieving millisecond-level inversion speed. We compared the results of different shape models using the Chamfer distance between traditional methods and ours and found that our method performs better, especially when handling special shapes. For the detection of concave areas on the convex hull, the intersection over union (IoU) of our predictions reached 0.89. We further validated this method using observational data from the Lowell Observatory to predict the convex shapes of the asteroids 3337 Miloš and 1289 Kutaïssi, and we conducted light curve fitting experiments. The experimental results demonstrated the robustness and adaptability of the method. Conclusions. We propose a deep learning-based method for asteroid shape inversion using light curve data to reconstruct the convex hull of asteroids and predict concave areas on the convex hull of non-convex asteroids. Our deep learning model efficiently extracts features from input data through convolutional and transformer networks, learning the complex illumination relationships embedded in the light curve data, and enabling precise estimation of the three-dimensional point cloud representing asteroid shapes.