Computer-Vision-Based Structure Shape Monitoring of Bridges Using Natural Texture Feature Tracking

Abstract

Structural health monitoring is carried out for a limited number of measuring points of the bridge. The root of the problem of bridge damage identification is that the mechanical equation inversion result is not unique due to the incomplete measured data. The full-field description ability of digital images to the structural shapes can effectively alleviate the problem of damage identification caused by incomplete measured data. This project aims to research the full-field shape monitoring method of the bridge. Firstly, the mathematical representation of the corresponding image points on the bridge surface is formed by using the image feature extraction method. Analyze the feature points position change mathematical model before and after deformation, and propose the structural full-field displacement vector calculation theory. Verify the full-field displacement calculation theory by a test beam. The results show that the maximum absolute error of the vector length is 0.24 mm and the relative error is less than 5%. The research realizes the structural full-field displacement monitoring under the natural texture condition for the first time. The results can promote the application and development of digital image processing technology in the field of structural health monitoring, improve the level of bridge safety evaluation, and realize the automation, intelligence and visualization of structural deformation monitoring.