A Noval Building Vibration Measurement system based on Computer Vision Algorithms

Vibration measurement and analysis can provide data support for structural health monitoring, which is of great significance in building structure inspection. Conventional vibration detection methods have to arrange acceleration sensors within the structure, which can be invasive and inconvenient. In contrast, video motion amplification algorithm can quantitatively measure the structural vibration information based on video monitoring, realizing long-term and long-distance measurement. This paper established a systematic procedure of building vibration detection based on video monitoring. The system includes motion amplification, mask detection, and vibration signal extraction, combining video cameras and edge computing devices. Using the collected video as the system inputs, one can obtain the basic characteristics of the building vibrations such as deformation amplitude and vibration frequency. As a validation, the proposed method was compared with an infrared amplitude measuring instrument. The results show that the accuracy of vibration frequency reaches 99.8%, and the accuracy of amplitude reaches 93%. The procedure was then applied to the vibration acquisition of a 94m residential building. The proposed method successfully realized long-term and real-time measurement of the building vibrations, and it can also guarantee the same level of accuracy as that measured by traditional methods.

I. Introduction

With the increasing demand for refined measurement in industrial product inspection, civil structure evaluation, biological monitoring, surgical operation, and entertainment industries in recent years, there are more and more scenarios that require visual processing of tiny vibrations. For example, in the medical field, the tiny vibrations of neurovascular bundles that are difficult to be observed in endoscopes can be magnified to assist doctors in performing operations and improve the success rate of operations [1]. The amplification and analysis of facial micro-expressions can be applied to the analysis of audience emotions during movie viewing, and can also be used to distinguish whether the video is faked by digital technology [2],[13],[15]. The application scenarios of motion amplification are becoming more and more extensive.