Astro-Det: Resident Space Object Detection for Space Situational Awareness

The flourishing space industry has caused a significant increase in space debris, posing a critical risk to high-asset operational spacecraft. A particular challenge is identifying small-sized debris characterized by low reflectivity and rapid motion. Traditional ground-based detection methods, constrained by geographic and daylight limitations, necessitate an exploration of in-space debris identification to bolster spacecraft’s space situational awareness (SSA) capability. Onboard optical sensors, which generate high-resolution images, offer crucial data for these tasks. Considering the sparsity and low signal-to-noise ratio (SNR) in space scenario images, this paper introduces an Astro-Det framework to detect streaks and aggregated spots in sparse environments. Within this framework, morphological and learning-based methodologies are independently applied to assess the effectiveness of object detection in space. A novel onboard camera-based space dataset is leveraged to evaluate the framework. Simulation results reveal that while the morphological approach falls short in detection accuracy compared to the learning-based approach, it demonstrates superior computational efficiency. This research presents detailed comparative evaluation, providing valuable insights into how these established algorithms perform under the unique conditions of sparse space scenarios.