Aiming at the problems such as complex environment, low separation rate and difficult to identify the targets in the process of underwater sonar target search, a new method based on dual tree double density complex wavelet subband model is proposed to identify the underwater target automatically.Firstly, the dual tree double density complex wavelet is used to decompose the acoustic image to get different direction subbands.Then,usingthe three parametersgeneralizesΓ function of different directional subband coefficients to fit as feature items.Finally, usingthe difference in distribution characteristicsbetween acoustic target image and background sonarimage calculating the KL distance,the similarityis matchedandthe optimal threshold is setto distinguish the existingtarget automatically.The experiments prove that the proposed methodis applied to the identification of sonar images with multiple targets,the quality factor of detection is 97.2% and the detection quality factor is 10% higher than that of other methods at different noise levels,and has high detection rate and robustness.