Abstract: Traditional detection of broadband targets in passive sonars has low output signal-to-noise ratio and poor performance in a complex situation with multiple targets and strong interferences. To solve this problem, a target detection method is proposed based on the characteristics of the energy distribution of broadband signals in the frequency-wavenumber domain by using uniform linear array. The proposed method converts the array signal into the frequency-wavenumber domain and uses the characteristics of the width and the spatial distribution of the main lobes and the side lobes to discriminate the main lobes in the wavenumber domain. After discriminating the main lobes of the same target at different frequencies, the accumulation of main lobe energy and the number of main lobes are used as the azimuth spectra for target detection. The theoretical analysis and simulations show the proposed method only utilizes the main lobes which have prominent contributions to target detection, thereby reducing the influence of the side lobes dramatically and improving the detection performance significantly. The results of trial data processing show that the output signal-to-noise ratio of the proposed method can be increased by 5.58 dB compared to SPED and 8.73 dB compared to CED. In addition, the computing time is decreased by 46% compared to CED, which validates the superiority of the proposed method.