Research On Non-contact Identification Algorithm And Application Of Bridge Structure Dynamic Displacement Based On High Frame Rate Video

In the long-term service of the bridge structure,due to environmental erosion,cyclic load and other effects,the bearing capacity and durability are constantly reduced,which in turn affects the safety of road and bridge operations.The dynamic parameters of the bridge structure are important indicators for the macro evaluation of the overall stiffness and service performance of the bridge structure,and also a key scale for evaluating the safe operation capability of the bridge.The traditional testing method requires a lot of manpower and material resources to install the sensor device.At the same time,the sensor is susceptible to resonance caused by the traffic and produce resonance.Especially in the complex and harsh environment,this contact sensing method makes the field implementation more difficult.In recent years,with the development and rapid popularization of modern communication equipment,structural dynamic parameter identification based on machine vision plays an increasingly important role in the operation,maintenance,management and maintenance of bridge structures.Aiming at the problems of low recognition efficiency,omission of logo points and misalignment in the recognition process of commercial algorithms,the hardware requirements of the imaging system are high in the long-range weak amplitude vibration test environment,and the imaging system susceptible to environmental vibration during the test,this paper has proposed a non-contact identification method for bridge dynamic characteristics based on high-frame video analysis.The main research contents are as follows:(1)Combining morphological processing and image distortion repairing algorithms,the image defects and distortion distortion of vibration test video image are preprocessed and repaired.The algorithm of connected region recognition based on geometric feature extraction is proposed to realize the dynamic location and tracking of the markers in the process of bridge dynamic test.This method has the characteristics of fast and accurate detection,which can accurately identify the mark points installed on the surface of the structure with special geometry,filter all non target points,and obtain the structural vibration time history information.At the same time,the algorithm has fast operation speed and high efficiency,which is suitable for bridge power testing with huge data volume,saving the identification workload,and avoiding the omission and misalignment in the identification process of other commercial algorithms.The algorithm is applied to the free vibration test of simply supported steel beams in the laboratory environment,the human-induced vibration test of the cable-stayed pedestrian bridge in the outdoor environment,and the vehicle-induced vibration test of the long-span cable-stayed bridge in the bay environment.Compared with the test results of acceleration sensor,which proves that the algorithm can achieve accurate positioning and efficient tracking of bridge structure under various test conditions.(2)The super-resolution reconstruction algorithm based on deep learning is introduced into the non-contact measurement of bridge dynamic characteristics.Taking the experimental video image of this paper as the training sample,the super-resolution reconstruction algorithm model based on the structural vibration test image is obtained through deep learning,the learned model was used to super-resolution restoration of experimental images under different working conditions.When the lens focal length and camera pixels are limited,the obtained video image is reconstructed and upgraded at the algorithm level,and the quality of the reconstructed image is evaluated based on the peak signal-to-noise ratio.The results show that,for most of the experimental images of the working conditions,the super-resolution resolution image quality is very good,which proves the feasibility and practicability of the super-resolution reconstruction algorithm in the bridge vibration experiment.Then it is applied to the free vibration test of simply supported steel beam in laboratory environment and the human induced vibration test of cable-stayed footbridge in outdoor environment,which improves the resolution and clarity of video image,significantly improves the accuracy of time history curve,the order and accuracy of spectrum analysis results,and reduces the dependence on the hardware facilities of imaging system.(3)Aiming at the impact of external environment vibrations such as bridge deck wind on the stability of the imaging system itself,the empirical mode decomposition algorithm is used to decompose the spectrum results obtained from the test.Based on the vibration test of the long-span cable-stayed bridge vehicle in the bay environment,the noise components generated by the environmental vibration are eliminated,and the influence of the environmental vibration on the imaging system is significantly reduced.The applicability of the connected domain recognition algorithm based on geometric feature extraction proposed in this paper is improved in the testing of long-span bridges in the bay environment.