Abstract: Radar has become one of the most important tools in the field of bridge health detection, mainly because of its non-contact, continuous, comprehensive detection and other advantages. Radar detection of bridge health mainly through the analysis of bridge vibration signals, radar vibration signals contain phase information, easy to extract shape variables. For the detection of vibration signals, it is usually necessary to carry out deformation inversion of the original signal and then detect the vibration signal on the inversion signal, but the bridge vibration signal is usually in a complex environment. Under the influence of clutter and noise, the inversion signal is difficult to show the characteristics of vibration signal itself. hence, this paper proposes a bridge frequency detection method based on CA-CFAR, which first detects the distance element in the distance doppler domain, and then detects the signal frequency based on the characteristic that the vibration signal itself has periodicity and multiple harmonics. the method in this paper directly detects the frequency of radar signal and reduces the error caused by the inversion of shape variables. First, the geometric model of vibration signal and the signal model are given. Then the principle of detection algorithm is discussed. Finally, the feasibility of the algorithm is verified by experiments.