Dynamic displacement recognition of frame structures based on computer vision

Abstract

Displacement induced by external forces is one of the most intuitive variables for assessing structural safety. Traditional contact methods, such as deploying displacement sensors on structures, are often limited by objective factors. Non-contact methods, such as utilizing computer vision algorithms like optical flow estimation and feature matching, offer the advantages of rapid and accurate structural displacement acquisition, unaffected by the structure itself and quick deployment. However, enhancing the accuracy of displacement monitoring based on computer vision remains a focal point in this field of study. In this paper, based on experimental data from a vibration table testing a four-layer reinforced concrete framework under three different conditions, we propose a method for processing dynamic displacement data that combines non-contact and contact approaches. This method integrates dynamically recognized structural displacements based on computer vision technology with data recorded by acceleration sensors on the structure to enhance displacement monitoring accuracy. The research results demonstrate that our method can obtain structural dynamic displacements based on computer vision information, confirming the effectiveness and reliability of our approach.

Full Paper