Missing area detection and damage mapping method based on image coordinates for bridge inspection using unmanned aerial vehicles

Abstract

Safety inspection and maintenance of infrastructure including bridges improve the reliability of structures and prevent unexpected accidents due to the aging of facilities. The conventional bridge inspection method have several limitations that cost and time are inefficient, and inspectors are exposed to danger during inspection. In addition to the limitation of the conventional human-based bridge inspection, the development of image sensors and deep learning technology is leading to bridge inspection using Unmanned Aerial Vehicle (UAV). However, there are limitations that need to be addressed for the practical application of UAV-based bridge inspection. Among them, conventional methods using UAV are insufficient to identify missing areas from acquired images and map the detected damage based on deep learning. In order to overcome the limitations of practical application in UAV-based bridge inspection, this thesis proposed image coordinate-based missing area detection and damage mapping method using metadata of the UAV system. The methodology of missing area detection and damage mapping consists of 4 phases. Phase 1 is a procedure to determine the coordinate of the center position of the camera from the location information of the GPS sensor mounted on the UAV. In this phase, the spatial coordinate transformation from the global coordinate system to the UAV coordinate system is performed. In phase 2, the center point coordinates of each image are estimated using the determined center position coordinate of the camera and distance information between the camera and the target object. For phase 3, the coordinates of each image are estimated by calculating the field of view (FOV) size using the working distance and the focal length of the camera. In the final phase, the method of the missing area detection and the damage mapping based on image coordinates of each image is described. The proposed image-coordinate-based approach is validated by using a real-size grid board, and the experimental validation of the missing area detection and damage mapping method is conducted on the concrete shear wall and the actual bridge. The detected missing areas and damage mapping results are compared with image stitching and human-based inspection results, respectively. As a result of experimental validation, the proposed methodology provided the results of the detection of missing areas and mapping of detected damage within appropriate accuracy.