用合成几何监督提升单目航天器6D姿态估计精度。
GAP-GDRNet: Geometry-aware monocular 6D pose estimation for spacecraft using synthetic geometric supervision
- 引入全局结构注意力与局部弱纹理增强，优化几何回归流程。
- 旋转误差1.96°，平移误差0.0165m，[email protected]达95.16%。
- 适用于弱纹理、遮挡等挑战场景，适合航天器视觉定位任务。
单目航天器6D姿态估计在纹理弱、结构细、光照变化和遮挡条件下仍具挑战。本文提出GAP-GDRNet，基于GDR-Net构建的几何感知RGB框架，针对单目标合成航天器数据集设计。方法在两点强化几何引导回归：一是在密集几何预测前加入AFR模块，融合全局结构注意力与局部弱纹理增强；二是在Patch-PnP中插入PGSA，关联下采样几何区域以辅助最终姿态回归。通过Blender渲染与标注流程生成掩码、模型坐标图、相机内参及6D姿态标签，实现密集监督。在自建航天器数据集上，该方法旋转误差1.96°，平移误差0.0165米，[email protected]米达95.16%，相较复现的GDR-Net基线提升3.88个百分点，推理速度达35.97 FPS。在T-LESS和LM-O上的测试也显示其对无纹理、遮挡的非航天器物体具有持续优势。
原文摘要 · Abstract (English)
Monocular spacecraft 6D pose estimation remains difficult under weak texture, thin structures, illumination variation, and occlusion. This article presents GAP-GDRNet, a geometry-aware RGB framework built on GDR-Net for a single-target synthetic spacecraft benchmark. The method strengthens the geometry-guided regression pipeline at two points. First, AFR is placed before dense geometric prediction to combine global structural attention with local weak-texture enhancement. Second, PGSA is inserted into Patch-PnP to relate downsampled geometric regions before final pose regression. Dense supervision is obtained from a Blender-based rendering and annotation process that provides masks, model-coordinate maps, camera intrinsics, and 6D pose labels. On the self-built spacecraft dataset, GAP-GDRNet achieves a rotation error of 1.96°, a translation error of 0.0165 m,and 95.16% [email protected] m, outperforming the reproduced GDR-Net baseline by 3.88 percentage points while running at 35.97 FPS. Tests on T-LESS and LM-O further show consistent gains over the reproduced baseline on textureless and occluded non-spacecraft objects.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。