Markerless LED-based geometric calibration for high-precision image acquisition in structural inspection

This study proposes an image acquisition system that aims to correct geometric distortion of images for the structure inspection. The proposed system is composed of main camera, l􀵴ght-em􀵴tt􀵴ng d􀵴ode module, the displacement sensor, and the movement system. Most images taken in the industrial environment are typically distorted, which increase difficulty of damage inspection. Thus, calibration procedure is required to correct distortion. However, two main problems remain: (1) most existing studies focus on lens distortion without considering geometric distortions caused by camera angles, and (2) most image acquisition systems rely on markers such as checkerboard, which increase inconvenience. Therefore, the proposed image acquisition system is developed to ease these problems. The proposed system employs an LED module to project a specialized light pattern instead of using physical markers, and geometric calibration can be performed by selecting corner points of the projected pattern. Furthermore, the movement system supports linear movement up to 3000 mm and full 360° rotation, producing flexibility in experimental setups. Thus, experts are able to perform precise inspection with simple image preprocessing. Experiments are conducted in both indoor and outdoor environments, and results prove the effectiveness of the proposed image acquisition system by achieving better root mean square error.