Development of Marker-Free Night-Vision Displacement Sensor System by Using Image Convex Hull Optimization.

Vision-based displacement sensors (VDSs) have the potential to be widely used in the structural health monitoring field, because the VDSs are generally easier to install and have higher applicability to the existing structures compared to the other conventional displacement sensors. However, the VDS also has disadvantages, in that ancillary markers are needed for extracting displacement data and data reliability is significantly lowered at night. In this study, a night vision displacement sensor (NVDS) was proposed to overcome the aforementioned two limitations. First, a non-contact NVDS system is developed with the installation of the infrared (IR) pass filter. Since it utilizes the wavelength of the infrared region and it is not sensitive to the change of a visible ray, it can precisely extract the shape information of the structure even at night. Second, a technique to extract the feature points from the images without any ancillary marker was formulated through an image convex hull optimization. Finally, the experimental tests of a three-story scaled model were performed to investigate the effectiveness of proposed NVDS at night. The results demonstrate that the NVDS has sufficiently high accuracy even at night and it can precisely measure the dynamic characteristics such as mode shapes and natural frequencies of the structure. The proposed system and formulation would extend the applicability of vision sensor not only into night-time measure but also marker-free measure.