提出了一种基于多频涡流信号处理的结构识别方法，该方法首先采用多频涡流仪获得低频绝对通道信号，通过旋转变换得到多组信号对，确定振幅最小的那一对，接着运用中值滤波技术获得整体信号幅度摆动趋势线，从而消除信号摆动影响，凸显结构信号；然后，根据结构幅度阈值和最小间距确定结构中心位置及其数量；最后，根据弯管和支撑信号特征，识别并显示这些结构。运用多组来自不同核电站的实测数据进行识别，结果显示了该方法的有效性。

In this paper, a method for structure recognition based on multi-frequency eddy current signal processing is put forward, in which low-frequency signals in absolute channel obtained from multi-frequency eddy current instrument are adopted, and several pairs of signals are obtained by revolving transformation, and that pair with least amplitude is determined. The trend of the whole signal is got by median filtering and is used to cancel the sway and stand out the structure signals. And then, the structure centers and their number are determined by the amplitude threshold and the least interval. At last, these structures are recognized and shown based on the characters of bend tubes and supports. The effectiveness of this algorithm is proven by recognition of multiple groups of data from several NPP.