An epipolar constraint-based method to compensate for camera motion in monocular vision-based asphalt pavement deflection measurement

In-situ measurements of surface deflection are essential for assessing the structural capacity of in-service pavements. Conventional techniques rely on multiple laser sensors, which increases system complexity and cost. In contrast, monocular vision-based techniques offer a promising alternative due to their ability to measure full-field deformation with just a single camera. However, external disturbances can induce camera motion, introducing significant systematic errors. Since asphalt pavement deflections are typically sub-millimeter, sometimes below 0.10 mm, high-precision motion compensation is crucial. However, limitations such as the inability to use artificial markers and pavement surface irregularities prevent existing methods from achieving the required accuracy. To address this challenge, this paper proposes a new camera motion compensation method based on the epipolar constraint. First, a single camera sequentially captures the reference and deformed images with relative poses approximating parallel binocular vision. Second, a pavement region assumed to be undeformed is selected to estimate the fundamental matrix, which defines the epipolar constraint used to compensate for camera motion. The accuracy and robustness of the proposed method are validated through simulation experiments using a physics-based graphics model (PBGM). Field tests further confirm its practical effectiveness. This method provides technical support for the development of a vehicle-mounted vision system for asphalt pavement deflection measurement.