As various structures rapidly develop in the engineering field, the measurement of large displacement structures has become an important research topic. Traditional measurement methods cannot meet the practical needs of multipoint or full-field measurements. The binocular stereo vision system has been widely applied in engineering as a contactless displacement measurement tool due to its advantages of simple optical path, full-field non-contact measurement, and large measurement field of view. However, existing binocular stereo vision measurement technologies typically require the binocular cameras to remain fixed after pre-calibration to maintain the stability of the reference coordinate system and the relative pose of the cameras, which ensures the effectiveness and correctness of 3D displacement calculation. This limitation also restricts the measurement range of the system to a certain extent and may cause measurement failures if the area of interest moves out of the field of view due to large deformations. Therefore, the development of a system with tracking measurement capabilities has become an urgent need to balance measurement range and accuracy. We aim to develop a novel technique that can continue measurement in such scenarios.

We propose a large displacement measurement method based on panning binocular stereo vision, which utilizes imaging units mounted on separate dual-axis panning stages to form a panning vision system. The intrinsic parameters of the imaging units are predetermined through calibration using a chessboard pattern. During large displacement measurements, external parameters are calibrated using coded markers based on the epipolar constraint relationship, followed by an inverse depth bundle adjustment. During measurement, the panning binocular stereo vision system adjusts as the target moves out of the predefined field of view. The external parameters are calibrated using the aforementioned method, and the current coordinates are transformed back to the original using the iterative closest point algorithm. By supplementing corner points or encoded points in the overlapping areas of the field of view before and after panning, a rigid body transformation matrix for the left camera is constructed based on spatial coordinates before and after panning, which ensures alignment with the global coordinate system. Finally, displacement information for the measurement point is calculated during the entire motion.

Three experiments are presented in this study. Firstly, we conduct the proposed camera calibration method, showing that the measured displacement value closely matches the real displacement (Fig. 6). With a reference 3 m horizontal field of view, the relative errors in the in-plane and off-plane directions are 0.040 and 0.047 mm/m, respectively. Secondly, we evaluate the accuracy of coordinate transformation based on ICP, with experimental results indicating a relative error of 0.11 mm/m for the coordinate transformation. Finally, we apply the developed panning binocular stereo visual system to measure large displacements. We use a one-dimensional displacement stage to simulate target movement beyond the field of view, which enables continuous displacement measurement. With a single panning operation, the field of view expands from 3 m by 3 m to 5.5 m by 3 m, and the relative errors in predefined translational displacement are less than 0.3% (Fig. 10).

A measurement method based on panning binocular stereo vision has been proposed to facilitate measurements when the target moves out of the field of view. This method combines fast camera calibration with key technologies in coordinate transformation. The binocular imaging system adjusts flexibly based on the positional information of the target point within the field of view, effectively preventing measurement failures due to target loss during large displacements. To optimize the acquisition of reliable external parameters, a refined strategy is introduced for camera calibration and coordinate transformation, which integrates optical coding markers and inverse depth bundle adjustment. For fixed feature points within the field of view before and after panning, the ICP algorithm is used to establish the coordinate transformation relationship and complete the entire motion measurement of the target point. This developed system expands the applications of existing binocular stereo vision with accurate displacement measurements, thereby offering a practical and effective solution for measuring large displacements in engineering structures.