Three-dimensional monitoring of non-rigid ground surface deformation based on time-series multi-view collaborative tracking

Abstract High spatiotemporal resolution surface deformation information offers the advantage of comprehensively, precisely, and rapidly revealing surface deformation characteristics, and is key data for early identification and dynamic warning of geological disasters such as landslides and ground subsidence. However, the vast majority of monitoring methods suffer from shortcomings such as low spatiotemporal resolution, and traditional feature tracking algorithms exhibit limited reliability in long-term, non-rigid deformation scenarios. To address these issues, a surface deformation tracking method based on Time-Series Multi-View Collaborative Tracking (TS-MVCT) was constructed. The TS-MVCT framework estimates the spatial positions of the initial point cloud after deformation by establishing 2D correspondences between temporally adjacent observations within consecutive short-term intervals, followed by reliability-weighted depth fusion for 3D position estimation. To validate the effectiveness of the constructed method, an indoor simulated surface deformation experiment was conducted. The experimental results demonstrate that, compared to the M3C2 algorithm, which primarily characterizes displacement along the local surface normal, TS-MVCT effectively reduces deformation underestimation and the loss of displacement components in different spatial directions, while enabling the estimation of 3D displacement vectors at a higher density than SIFT-FT. Quantitative validation under laboratory conditions indicates that the relative deviations of the retrieved 3D displacements—measured against manually annotated reference values—are generally within 2 mm. On markerless natural surfaces, the algorithm achieves PSR and reprojection-error values within a similar numerical range to those of artificial-target regions, indicating comparable internal matching and multi-view geometric consistency under the controlled experimental conditions. In summary, the TS-MVCT framework presents a viable non-contact approach for acquiring 3D surface deformation information, providing deformation data for future geological hazard monitoring.