Abstract Measuring the vibration frequency of structures with image processing methods is one of the non-contact measurement methods. However, the frequency to be measured should be lower than the camera capture speed. In this study, a method was developed to extract and to compare pattern characteristics of data acquired from a pixel of a low-speed camera. At the first stage, images of a rotating disk were recorded at known frequencies to find out the camera’s response. After that, distinctive features were extracted both from the raw and clustered intensity data acquired from the pixels of the recorded images. At the second stage, pixel intensity data of a rotating disk with an unknown frequency was also recorded with the same camera. Similarly, features were extracted from these data, and they were compared with the features at known frequencies. A frequency range was found for the unknown frequency value in this way. Finally, the unknown frequency value was estimated by interpolation in the frequency range. The method was verified on an experimental setup, and five unknown frequency values were successfully estimated. Consequently, higher frequencies than the capture speed of the camera were correctly estimated using the developed method. On the other hand, it was also noticed that the clustering of the discrete intensity data was also significant.

中文翻译：

---

一种新算法的开发和实验验证，以估计比相机捕获速度更快的物体移动速度

摘要 用图像处理方法测量结构振动频率是非接触测量方法之一。但是，要测量的频率应低于相机捕获速度。在这项研究中，开发了一种方法来提取和比较从低速相机的像素获取的数据的模式特征。在第一阶段，以已知频率记录旋转磁盘的图像以找出相机的响应。之后，从从记录图像的像素获取的原始和聚类强度数据中提取独特的特征。在第二阶段，一个未知频率的旋转圆盘的像素强度数据也用同一台相机记录下来。同样，从这些数据中提取特征，并将它们与已知频率的特征进行比较。以这种方式为未知频率值找到了频率范围。最后，通过在频率范围内插值估计未知频率值。该方法在实验装置上得到验证，并成功估计了五个未知频率值。因此，使用开发的方法正确估计了比相机捕获速度更高的频率。另一方面，还注意到离散强度数据的聚类也很重要。因此，使用开发的方法正确估计了比相机捕获速度更高的频率。另一方面，还注意到离散强度数据的聚类也很重要。因此，使用开发的方法正确估计了比相机捕获速度更高的频率。另一方面，还注意到离散强度数据的聚类也很重要。