Event‐Based Attitude‐Driven Functional Damage Assessment of Space Targets

ABSTRACT

Functional damage to space targets often appears as subtle attitude instability rather than visible destruction, making reliable assessment difficult for conventional image‐based methods. These methods are sensitive to illumination, viewing geometry and prior assumptions and often miss weak dynamic signatures related to damage. To address this issue, this paper proposes an event‐based functional damage assessment framework based on high‐precision attitude variation estimation. By exploiting the asynchronous sensing mechanism, high temporal resolution and wide dynamic range of event cameras, the framework captures fine‐grained motion responses under complex illumination and high‐dynamic conditions. A hybrid dataset combining simulated rendering and real event‐camera measurements is constructed for training. Based on reconstructed event frames, a joint architecture integrating Faster R‐CNN, Feature Pyramid Network and High‐Resolution Network is used for target detection and keypoint regression. The predicted two‐dimensional keypoints are matched with predefined three‐dimensional keypoints of the target model and then used in a Perspective‐n‐Point algorithm to estimate pose and derive attitude variation parameters. A noise‐aware denoising model and an adaptive integration strategy based on the local surface of active events are further introduced. Experiments show a mean attitude estimation error below 0.15°, enabling reliable functional damage classification.