Safety inspection and monitoring technique using real-time image processing from a single image for mechanically stabilized earth wall

This study applied an image processing method to inspect and monitor stability of mechanically stabilized earth wall (MSE). Public institutions regularly perform safety inspections, but they can not detect changes in facilities that occur in real-time. In addition, previous researches can not detect unspecified positions of the facility because a sensor is installed at a specific position. Safety inspection and monitoring technique using real-time image processing can solve spatio-temporal limits of existing methods. Most of displacement measurement technique were using more than 2 images to take different viewpoint and to reconstruct and analysis 3D point cloud, Therefore, it is not possible to use installed cameras such as CCTV, and there is a disadvantage that additional cameras, equipments and softwares must be newly installed. However, in this study, we try to detect the change of entire structure by extracting and analyzing video from a fixed point. The proposed technique consists of 5 steps : (1) Extracting images of MSE from video, (2) Correcting perspective distortion of images, (3) Extracting edge for image of MSE, (4) Detecting lines and circles by using Hough transform, and (5) Detecting displacement and rotation angle of each. The results show that the proposed method detects displacements of MSE and measuring errors were less than 10 mm. The proposed technique using image processing therefore can be widely used to inspect and monitor the stability of MSE.