Abstract: The interference of camera motion is an important cause of error in video vibration detection. To solve this problem, this paper proposes a cross-resistance sample consensus method against jitter and vibration which can effectively separate the vibration signal and camera motion signal in video, thus improving the reliability of video vibration detection. First, candidate feature points are extracted through SURF（speeded-up robust features） algorithm, and the mutual inhibition measure of vibration and camera motion is designed which can separate the candidate feature points to obtain the feature points of camera motion. Second, it registers the video images according to the feature points of camera motion to obtain a video sequence that removes the interference of camera motion. Finally, for the video sequence of the stabilized camera, Eulerian video vibration detection is used to obtain the vibration frequency. Videos under different camera movements are collected, and the parameters of mutual inhibition measures are estimated. The accuracy of the vibration frequency is better than the existing non-contact vibration detection methods through the verification of the test set data.

Key words: vision measurement, frequency detection, cross-resistance sample consensus method against jitter and vibration, image registration

摘要： 相机运动的干扰是造成视频振动检测误差的重要原因。针对该问题，提出一种互抑制一致采样方法，对视频中的振动信号和相机运动信号实现有效分离，从而提高视频振动检测的可靠性。通过SURF（加速稳健特征）算法提取候选的特征点，并设计了振动与相机运动的互抑制测度，对候选的特征点进行分离，以获得相机运动的特征点。根据相机运动特征点对视频图像进行配准，以获得去除相机运动干扰的视频序列。对稳定相机的视频序列，采用欧拉视频振动检测方法获得振动频率。自行采集了不同相机运动下的视频，并针对互抑制测度的参数进行估计。通过对测试集数据进行验证，得到的振动频率准确率优于现有的非接触振动检测方法。

关键词: 视觉测量, 频率检测, 互抑制一致采样算法, 图像配准