Motion blur compensated optical pose estimation for non-cooperative space targets using adaptive multi-scale ellipse fitting

Motion blur hampers optical navigation to non-cooperative spacecraft. We present an uncertainty-aware stereo imaging approach that restores blur and recovers relative pose with calibrated confidence. The method couples physics-guided deblurring with edge-preserving regularization, adaptive detection of elliptical structures on manufactured surfaces, and stereo geometric refinement that propagates measurement uncertainty to the final estimates. Comprehensive laboratory experiments on a space-like optical bench, supported by matched synthetic trials spanning mild, moderate, and severe blur, show lower position and orientation errors and higher success rates than five representative baselines, with the largest gains under severe blur, while keeping runtime practical. Sensitivity studies and coverage-confidence analyses confirm well-behaved uncertainty for risk-aware guidance and docking.