Development of a Robust Vision-Based Interstory Deformation Sensing Method Using a Kernelized Correlation Filter

In this study, we developed a vision-based interstory deformation sensing system. In vision-based vibration sensing, the uncertainty of measurement accuracy is a problem caused by the lack of robustness of image processing algorithms. We developed an image processing algorithm to improve robustness using a kernelized correlation filter (KCF). Evaluation of our KCF-based algorithm shows that the KCF-based algorithm is more robust than conventional image-processing methods. For example, when the extent of motion blur was large, the proposed algorithm was over 2.5 times more accurate than other algorithms in the reported literature. Our algorithm demonstrates advantages over other methods in allowing measurements in low-light environments and not requiring the use of a high-sensitivity camera.