Video Motion Amplification And Its Application In Modal Identification Of Structures

The non-contact measurement method which combines the optical measurement of digital camera with the visual image processing algorithm can solve many problems such as the additional mass of traditional sensor and the non-synchronization and packet loss of wireless sensor.It has become a new research direction of civil engineering structure and has been successfully used in the measurement and analysis of structural vibration.However,the traditional machine vision measurement method requires a large number of high-contrast markers or patterns to be arranged on the surface of the structure under test,which is timeconsuming.When the area to be measured is large or difficult to reach,this method is also slightly inadequate.In addition,the stereophotogrammetry system is limited by the inherent noise of image acquisition and camera resolution.For traditional optical measurement systems,the excitation of the structure under test with only a small amount of energy may produce subtle vibration response that is difficult to capture.This paper reviews some research results of machine vision motion amplification and its application in modal identification of engineering structures.In this paper,the main methods of traditional motion amplification techniques and their application frames in practical engineering are reviewed.In terms of motion amplification techniques,Euler linearity,Euler phase-based motion amplification and two kinds of supervised motion amplification techniques related to neural networks are introduced.The latest techniques in large motion isolation are introduced,which are based on layered motion amplification,depth perception motion amplification and Euler acceleration motion amplification.In addition,three kinds of motion signal extraction technologies are compared and analyzed: image correlation technology,point tracking technology and phase optical flow method.This paper focuses on the defects of the current machine vision motion magnification method in the practical application,and then carries out the corresponding research,and improves it in several aspects.First of all,based on the traditional motion amplification method too dependent on manual input filter or accurate a priori knowledge of target signal and the magnification coefficient of vulnerable to input corresponding pyramid space support limiting the two shortcomings,creatively and amplify the adaptive motion video into the process of the tiny changes in three key solution of the optimization problem.In order to effectively model the amplification transformation,the method of alternating direction multiplier is used to solve the convex variable detection optimization problem.After the transformation step,the forward additive iterative method is used to iteratively minimize the difference between the original video image sequence and the amplified sequence on the basis of enhancing the correlation coefficient,which solves the disturbance problem to some extent.In this paper,the method is compared with the most advanced technology through quantitative and qualitative structural microvibration experiments,and the visual quality of the amplified video is greatly improved.Secondly,machine visual motion amplification applied to engineering structures modal identification phase is a bigger problem status jump phenomenon,namely in the tiny motion detection or vibration signal extraction process,there will be an interval repeat step 2 aspect phenomenon,lead to the large video of artifacts,affect the accuracy of modal parameters identification.In this paper,the phase unwrapping method based on the brightness transfer equation is used to modify the original complex pyramid model.Firstly,the phase difference information obtained from the complex pyramid is initially processed,and then the iterative estimation is constantly updated,so as to carry out the adaptive phase unwinding and effectively eliminate the pseudo-image.Finally,this method is applied to the soft membrane structure and the actual pipeline vibration video,and the identification also verifies the effectiveness of this method and the robustness to the phase information.Finally,this paper summarizes the three future development directions of visual motion amplification :(1)improve the quantitative judgment criteria for different types of motion in the process of motion information extraction;(2)optimize the compatibility of the filter in the time domain and space domain;(3)improve the computational efficiency in the process of machine vision movement amplification.