Visual Vibrometry and its application to NDT through the use of various video and signal processing techniques

该项目侧重于视觉振动测量领域及其通过使用各种视频和信号处理技术在超声波无损检测（UNDT）中的应用。视觉测振法作为一种NDE方法的概念是通过利用普通消费者相机从标准帧率视频中恢复信号来探索的。视觉振动测量法有很大的潜力成为无损检测中快速大面积监测的主要方法。由于视觉振动测量的工作原理是基于使用中的损伤以人眼无法察觉的方式改变振动模式，但通过图像处理可以看到。无损检测的一种常见形式包括使用超声波来提取底层材料和系统的动态特性。该项目的一个目标是将超声波引导波的能力与视觉系统的能力结合起来，以便在深度穿透零件。特别是探索如何采用大面积的视觉监测方法来监测结构的状态。对于收集到的视频数据，采用强制激励，而不是简单地记录结构的被动响应。一个目标是成功地以明显大于相机广告帧率的频率对结构振动进行成像，并扩展相机的使用范围，超出其硬件规格。这可以通过利用UNDT中常见的短曝光时间和周期性激励来实现，即使激励频率超过奈奎斯特采样限制。这将涉及到使用频闪效果和利用相机曝光时间来实现这一目标，并从本质上“欺骗”相机系统以比正常情况下更高的帧率进行采样。这些情况的分析要考虑频闪效应，其中固有频率和输入激励分别是可计算的或已知的。最后的挑战将是，这些发展是否能够推进到足够远，从而允许从有用的超声波频率测量位移。为视觉振动法监测和检测零件损伤提供依据。为了实现这些目标，模拟和后处理工具已经开发出来，作为验证实验测试和预测实验测试行为的手段。其中，模拟视频旨在模仿真实数据，并使用根据实验测试结果定义的噪声模型来研究噪声与强度等级之间的关系，如图1所示。模拟测试视频提供了在物理测试之前测试实验装置的手段，并可用于探索关键摄像机参数对结果的影响。由于相机的动力学对测试性能有很大影响，因此将对相机的每个测试参数进行研究和量化。需要图像质量，帧率和曝光时间之间的关系，并且可能通过使用信噪比关系以及识别输出位移幅度和相位的准确性来量化。通过模拟探索每个摄像机和测试参数的影响，可以探索测试的局限性。例如，识别可能可靠、准确地从视频数据中提取的最小像素位移。

The project focuses on the field of visual vibrometry and its application to ultrasonic non-destructive testing (UNDT) through the use of various video and signal processing techniques. The concept of visual vibrometry as a method of NDE is explored by utilising regular consumer cameras to recover signals from standard frame-rate videos. Visual vibrometry has significant potential to be a major method for rapid large area monitoring within NDE. As visual vibrometry works on the basis that in-service damage alters vibrational modes in a way that is imperceptible to the human eye, but visible through image processing. A common form of NDT involves the use of ultrasonic sound waves to extract underlying material and dynamic properties of a system. One aim of the project is to combine the ability of ultrasonic guided waves, in order to penetrate a part at depth, with the ability of visual systems. In particular, to explore how large area visual monitoring approaches may be employed to monitor the state of structures. For the video data collected, forced excitation is utilised as opposed to simply recording the passive response of the structure One goal is to successfully image structural vibrations at frequencies significantly greater than the advertised frame-rate of the camera and extend the use of the camera beyond its hardware specification. This may be achieved by exploiting short exposure times and periodic excitation common in UNDT, even when the excitation frequency exceeds the Nyquist sampling limit. This will involve using stroboscopic effects and exploiting camera exposure times in order to achieve this and essentially 'cheat' the camera system in to sampling at a significantly higher frame-rate than normally possibly. These cases are intended to be analysed by taking into account stroboscopic effects produced, where natural frequencies and input excitations are either calculable or known, respectively. The final challenge will be whether these developments may be pushed far enough to allow the measurement of displacement from useful ultrasonic frequency. In order to provide the basis for visual vibrometry as a method to monitor and detect damage in parts. To achieve these goals a simulation and post processing tool has been developed as a means to validate experimental testing and to predict the behaviour of experimental testing. Where, the simulated video was designed to imitate real data and used a noise model defined on results from experimental testing performed to investigate the relationship between noise and intensity levels as shown in figure 1. The simulated test video provides a means to test experimental set ups prior to physical testing and may be used to explore the effects of key camera parameters on the results. As the dynamics of the camera are highly influential on the performance of the testing, each of the camera testing parameters will be investigated and quantified. A relationship between image quality, frame-rate and exposure time is required and will likely be quantified by using a signal to noise ratio relationship as well as the accuracy of the identified output displacement amplitude and phase. By exploring the effect of each camera and test parameter using the simulation, the limitations of the testing may be explored. Such as, identifying the smallest possible pixel displacement that may be reliably and accurately extracted from video data.