Abstract
Abstract
On-orbit photogrammetry of large deployable satellite antennas is challenged by intense solar stray-light interference, weak observation-network geometry, and the propagation of local measurement errors. To address these interconnected challenges, this paper proposes a robust photogrammetric framework tailored for harsh illumination environments. First, to mitigate the edge degradation and feature obscuration that small directional retro-reflective targets often suffer under strong background glare, we engineered an enhanced target equipped with a widened, low-reflectance black rim. At the algorithmic level, a top-hat transform is utilized to suppress the low-frequency, high-intensity background. By coupling Canny edge detection with gray-weighted centroiding, this approach achieves stable subpixel localization of target centers under severe stray-light interference. Furthermore, the weak spatial intersection geometry typical of on-orbit systems often causes local front-end errors to be nonlinearly amplified during bundle adjustment. Leveraging the unique structural characteristics of the FY-4M deployable antenna, we introduce a structure-constrained optimization strategy based on the local rigidity prior of individual antenna petals. Specifically, the relative geometric constancy of targets within a single petal substructure is formulated as a soft-constraint penalty term and integrated into the bundle adjustment model. This integration improves the stability and precision of 3D reconstruction under degraded network geometries. Validation through physical experiments and simulations shows that the proposed framework improves the robustness of target recognition and centroid extraction under severe stray light. It also improves the 3D measurement accuracy under weak geometric conditions. These results indicate that the proposed framework provides a feasible technical route for robust non-contact measurement of large deployable space structures.