The uncertainty analysis of bridge displacement in computer vision

Computer vision is commonly employed to measure the displacement of bridge structures. However, changes in environments often lead to poor data reproducibility. The uncertainty is an important parameter associated with the results, representing the dispersion of the data. This paper mainly discusses the uncertainty of measurement utilizing computer vision. The fundamental formula of measurement is derived from the principle of the pinhole camera model, followed by calculations of the uncertainty as well as a detailed analysis and deduction of the components that affect the uncertainty of measurement (e.g. pitch angle, deviation angle, observation distance, temperature, humidity). Both laboratory and on-site experiments have shown that the uncertainty model is more applicable than error estimation. In addition, the variation law of the uncertainty is revealed under the combination of different angles, laying a solid foundation for accurate measurement in the future.