计入字符识别跟踪的双目摄影振动测量方法

Binocular photogrammetric vibration measurement method introducing character recognition and tracking

针对大尺寸结构振动和模态测量要求,提出一种基于字符编码标志的双目摄影目标追踪振动测量方法。设计了一种由字符和圆形组成的编码标志,采用自适应维纳滤波进行图像去噪,自适应局部阈值法提取字符特征,基于BP神经网络训练进行字符识别,实现目标的追踪匹配;进行标志区域灰度平滑,采用灰度质心法获取标志中心坐标;基于双目摄影振动测量方法计算测点的时域振动响应。以某3kW风力机叶片为对象进行实验研究,结果与锤击法的前3阶频率相对误差均小于0.9%,表明该方法可靠可行。

Coded targets are widely used in photogrammetry for large-scale structures.A coded target based on a circled character is designed and applied in vibration measurement through binocular vision system,in which English or Chinese characters are used for coding information and the circle is used for positioning.Weiner filter method is employed to de-noise image.Further,character feature is segmented using adaptive local threshold method.Character information is identified through BP neural network.Target position is calculated with gray value gravity center method.The vibration responses of the coded targets position are measured by the binocular vision system.A 3 kW wind turbine blade was tested with the presented method.The first three natural frequencies were identified and the relative errors were less than 0.9%compared with those from hammer impact experiment.The results show that the method is reliable and feasible.