Track-Leakage-Free Hold-Out Self-Validation for Photogrammetric Reconstruction: Protocol, Sensitivity, and Limits
Track-Leakage-Free Hold-Out Self-Validation for Photogrammetric Reconstruction: Protocol, Sensitivity, and Limits
用于摄影测量重建的无轨迹泄露留出法自验证：协议、敏感性与局限性
Abstract: Automated photogrammetric inspection emits metric measurements from a 3D reconstruction whose own correctness is normally unknown without an external survey. Can a reconstruction estimate its own reliability with no ground truth? 摘要： 自动化摄影测量检测通过 3D 重建输出度量测量结果，但在没有外部测量的情况下，其自身的准确性通常是未知的。重建模型能否在没有地面真值（Ground Truth）的情况下评估自身的可靠性？
We formalise a track-leakage-free hold-out protocol: a deterministic subset of images is withheld and each re-localised against only 3D points seen by at least two retained images — a track-level barrier so a view is never tested against structure it helped create — aggregated into an mAA confidence. 我们提出了一种“无轨迹泄露”（track-leakage-free）的留出法协议：确定性地预留一部分图像，并将每张图像仅针对至少被两张保留图像观测到的 3D 点进行重定位——这形成了一个轨迹层面的屏障，确保视图不会被其自身参与构建的结构所验证——最终将其聚合为 mAA 置信度。
Held-out observations still entered the bundle adjustment that built the trusted structure, so we call it track-leakage-free (a stricter re-mapping variant confirms this at 3/5 deg). 虽然被留出的观测数据仍参与了构建受信任结构的束调整（Bundle Adjustment），但我们将其称为“无轨迹泄露”（一种更严格的重映射变体在 3/5 度范围内证实了这一点）。
Across operational GNSS-referenced captures plus ETH3D, EuRoC and the IMC 2025 benchmark we report four findings. 通过对实际 GNSS 参考采集数据以及 ETH3D、EuRoC 和 IMC 2025 基准测试的研究，我们报告了四项发现。
(i) The protocol is computationally well-posed: good reconstructions score millidegree self-consistency. (i) 该协议在计算上是适定的：高质量的重建模型可获得毫度级的自一致性评分。
(ii) Thresholded self-consistency saturates and does not track absolute accuracy: confidence stays near 1.00 while true RTK error swings up to 14x within a capture, and the per-capture correlation is sign-unstable across five captures (-0.57 to +0.98; pooled 95% CI spanning zero), so it cannot gate accuracy. (ii) 阈值化的自一致性会趋于饱和，无法追踪绝对精度：在单次采集过程中，置信度保持在 1.00 附近，而真实的 RTK 误差波动可达 14 倍；且五次采集的单次相关性符号不稳定（从 -0.57 到 +0.98；合并后的 95% 置信区间跨越零），因此它无法作为精度的门控指标。
(iii) It flags gross failure only when the failure destroys internal consistency: a fragmenting model drops confidence, but a single self-consistent, globally-distorted model evades it — three of four captures gave one model wrong by 55-106 m at confidence 1.00. (iii) 它仅在故障破坏内部一致性时才会标记出严重错误：模型碎片化会导致置信度下降，但一个自洽但全局扭曲的模型却能逃过检测——在四次采集中的三次，模型出现了 55-106 米的偏差，但置信度仍为 1.00。
(iv) The same dichotomy holds on independent ETH3D and IMC 2025 ground truth. Track-leakage-free hold-out measures internal geometric consistency, not absolute accuracy: neither a substitute for control-point assessment nor a general gross-failure gate. We release the protocol and degradation harness. (iv) 同样的二分法在独立的 ETH3D 和 IMC 2025 地面真值上也成立。无轨迹泄露留出法衡量的是内部几何一致性，而非绝对精度：它既不能替代控制点评估，也不能作为通用的严重故障门控。我们已发布该协议及退化测试工具。