A projected laser-crosshair-aided moving-camera high-speed videogrammetry method for dynamic displacement measurement in shaking-table tests

Abstract High-speed videogrammetry can provide non-contact three-dimensional displacement measurements for shaking-table bridge tests, but fixed camera layouts are often constrained by occlusion, limited laboratory space, and unfavorable viewing angles. Moving cameras improve observation geometry, yet their platform motion introduces pseudo-displacements that are difficult to remove when the specimen occupies most of the image and stable background features are unavailable. This study develops a projected laser-crosshair-aided moving-camera high-speed videogrammetry method to solve this compensation problem. Before excitation, surveyed benchmarks and structural targets define the engineering datum and undeformed reference configuration. During acquisition, projected laser-crosshair centres placed in the useful measurement view act as active optical observations for estimating the time-varying pose perturbation of each moving camera. A coarse-to-fine extraction procedure combines target localization, halo suppression, adaptive foreground separation, probabilistic edge-line modelling, consensus clustering, and four-edge centroid estimation to obtain subpixel laser-crosshair centres. The moving binocular system is then formulated as a non-rigid stereo model in which the left and right camera pose trajectories are optimized independently but synchronously in the common engineering coordinate system using Huber robust reprojection loss and temporal smoothness. After pose compensation, structural markers are reconstructed relative to the undeformed configuration to obtain dynamic displacement. Validation shows a projected laser-crosshair centre RMSE of 0.47 pixels. Controlled moving-camera tests achieve component-wise displacement RMSEs below 0.60 mm, and a representative shaking-table bridge test gives RMSEs of 0.24, 0.75, and 0.36 mm in the X, Y, and Z directions relative to a fixed stereo reference. These results demonstrate that surveyed laser-crosshair observations embedded in the measurement view can decouple camera-platform motion from structural response when the spatial laser intersections remain sufficiently stable, enabling sub-millimeter to millimeter-level dynamic displacement reconstruction in restricted shaking-table environments where fixed-camera geometry and passive background references are inadequate.