Multi-view Target-Free Video Structural Motion Estimation: A Self-adaptive Co-calibration Model

Abstract

Target-free magnification-based vibration measurements can be used to compute the full-filed dynamic response of structures, which alleviate the need for structural surface preparation and can be implemented in an efficient and autonomous manner. However, up until now, such methods have limitations in the measurement range of motion that can be estimated, which fundamentally restricts their applicability. Moreover, they are only for one-dimensional (1D) or two-dimensional (2D) measurement. A rapid and accurate multi-view video estimation method is highly demanded. This paper presents a novel multi-view target-free video structural motion estimation method based on a self-adaptive co-calibration model. The model is derived from the maximum likelihood formulation with sparsity priors. Fast cosine transform with extension to phase correction operation across the levels of multi-scale pyramid is introduced. To apply it to three-dimensional (3D) video estimations, an iteratively reweighted method is employed to handle the lq-norm multi-view minimization problem. The proposed method is demonstrated in case studies considering analyses of video motion estimates for structural health monitoring (SHM). Compared to conventional methods, the proposed method provides more accurate measurement of structural motions in both time and frequency domains.