Method of extracting structural seismic response from blurred surveillance videos based on line matching technique

Surveillance cameras have emerged as a promising alternative for capturing earthquake videos to obtain structural dynamic response information. However, surveillance videos captured in earthquake environments often suffer from blurring due to the uncertainties in filming conditions, posing a significant challenge to extracting structural seismic response from these videos, as current mainstream methods are ineffective in extracting feature point information from blurred videos. This paper proposes a novel framework based on the Adaptive Line Descriptor Hierarchical Filtering (LDHF) algorithm, which can rapidly, automatically, and accurately extract feature point information to calculate structural seismic responses from blurred surveillance videos. Initially, line segment detection techniques are employed to extract line descriptor information. Subsequently, the LDHF algorithm is utilized for robust and automatic line matching, with feature tracking points determined through the intersection of line segments. Finally, the seismic response is obtained by solving the camera pose estimation problem. The proposed method has been validated through numerical simulations and shaking table tests. Additionally, to simulate the blurring possibilities of videos recorded by different surveillance cameras during earthquakes, the original recorded videos are processed with motion blur and Gaussian noise to verify the robustness of the proposed method in handling videos with varying degrees of blur. The results indicate that this method significantly outperforms existing methods, achieving sub-pixel level accuracy. Another key feature is its ability to rapidly and automatically extract structural seismic responses, which offers a distinct advantage in processing large volumes of surveillance videos in the post-earthquake phase.