Research on Intelligent Perception Positioning and Threat Level Evaluation Based on Multi-source Information Fusion

The significant increase in space debris has led to a dramatic escalation in collision risks for spacecraft, necessitating the urgent development of high-precision space debris situational awareness and collision threat assessment technologies. This paper focuses on space debris targets and proposes a target perception and localization method that integrates multi-source heterogeneous observation data. By combining physical mechanism modeling with machine learning algorithms, a hybrid collision threat level assessment mechanism is constructed. At the perception and localization level, a unified mathematical model for multi-satellite observation is established, effectively integrating three types of heterogeneous observation data: distance, angle, and velocity. Subsequently, a Levenberg-Marquardt (LM) algorithm based on Huber weighting and iterative re-optimization is designed, significantly enhancing localization accuracy and algorithm robustness under abnormal data conditions. At the threat assessment level, a hybrid decision-making framework is proposed, which integrates an improved physical collision probability calculation model with a random forest classifier. This framework comprehensively considers both collision probability estimation and the impact of potential kinetic collision consequences, achieving efficient and precise classification of debris threat levels. Simulation results demonstrate that the proposed localization algorithm significantly outperforms traditional least squares estimation and geometric analysis methods in terms of root mean square error (RMSE) for position estimation. Simultaneously, the threat level classification model exhibits high overall classification accuracy, with the random forest classifier demonstrating superior discriminative performance compared to logistic regression models. This research provides an effective technical solution for space-based object surveillance missions and autonomous spacecraft collision avoidance decision support.