Abstract: In order to accurately detect the click signal of dolphins in marine environment with complex noise, an automatic detection method based on multi-parameter constraints is proposed. Firstly, the original data are divided into frames, then further converted from time domain to frequency domain with FFT. After that, energy calculation is performed, and an energy threshold is set to extract the pulse signal of suspected targets. Secondly, basing on characteristics of the click signals appearing in time series and the regularity of pulse interval inside the click train, the click signal pulse train is selected from the suspected target signals by constraining the number of pulses in the click train, pulse interval and other parameters. Finally, the automatic detection result is output, so the position of the click train and the number of click in the click train are obtained. The method is verified with in situ measurements, and the results show that the average recall rate of the dolphin click signal can reach 90% and the average false detection rate is 3.6% without manual intervention. The automatic detection method proposed in this paper can support the realization of acoustic monitoring of marine mammals and provide technical support for the study of biological behavior of marine mammals.