Vision-Based Structural Displacement Measurement Using Siamese Network

Structural displacement measurement can provide essential information of the structure that could be used to assess the condition of the structures. Recent computer vision technologies have provided an opportunity to obtain a more efficient way to measure the structural displacement. However, the traditional vision-based displacement measurement methods have few limitations. The user has to manually select the feature and the region of interest from the image frame and also needs to adjust parameters. Furthermore, when tracking fails, the user must manually repeat the process with new a region of interest. This procedure costs a plenty of time and effort. To overcome these limitations, this paper introduces a deep learning-based displacement measurement approach using Siamese network. The proposed method can automatically select and track feature points in the structure. To validate the performance of this method, a simulation-based experiment was conducted. The response of the 6-story building model was generated by simulation, and the animation of this building was encoded into a video. The displacement of the building was extracted by the movement of a region of interest from its original location. The validation test showed that the proposed method can not only automate the displacement measurement process, but also can get substantial accuracy of the measurement.