Independent temporal measurement readiness audit

Pre-registered2026-10-30. Audit adjacent pairs of the three archived frames on each of six spent drives:12 pairs. Do not fit temporal depth masks or extrinsic parameters. Pin images, camera/OXTS timestamps, selected OXTS packets and available calibration paths. Record camera intervals and same-index camera/IMU timestamp offsets; these are availability/timing evidence, not validated motion interpolation.

Available official calibration is IMU→Velodyne→camera. That chain consumes the reference LiDAR-camera extrinsic being evaluated, so it is not an independent metric camera-motion source for a selection rule. Record absence of configured direct camera/IMU or metric-camera-trajectory artifacts in this local input set. OXTS IMU motion is not silently re-labeled as camera motion. No true extrinsics enter the image-pair algorithm, even if reference files exist on disk.

Image-only support test: ORB2000, Hamming cross-check, deterministic seed0/thread1, fixed essential-matrix RANSAC .999/1px using original camera intrinsics, recoverPose cheirality. Record matches, epipolar and cheirality counts, and occupied4x4 cells in each image. Necessary directional-support screen: >=30 cheirality inliers and

=6 cells in both. This is a prespecified engineering screen, not a claim that ORB matches measure true overlapping surface fraction or accurate camera motion. Recovered monocular translation has unknown metric scale. E failures/missing features remain explicit zero support. Do not drop or retune any pair.

If image support is inadequate or independent metric motion is missing, report that as a temporal-depth-check blocker. An independently validated stereo/visual odometry module would be new work with its own frozen tests; current SGBM depth is under evaluation and cannot be treated as independent metric motion truth. No downloads, new recordings, selection using target extrinsics, ICP rerun or SOTA.