The identification of direct-path recordings for direction-of-arrival (DOA) estimation has been recognized as a crucial factor in enhancing the localization robustness under reverberant environments, especially for the challenging cases involving multiple acoustic sources. In particular, the recently introduced spatial feature denoted relative harmonic coefficients (RHC) has been extensively investigated for DOA estimation due to a strong association with spatial cues. Building upon this, this paper presents a multi-source localization method that utilizes higher-order RHC vector to detect direct-path dominant (DPD) bins in the time-frequency (TF) domain. Specifically, DPD bins are effectively identified by applying an orthogonal projection matrix (OPM) to the RHC vector. Moreover, we integrate the OPM with a closed-form single source DOA estimator, thus circumventing the exhaustive search over a dense spatial grid. Extensive experimental evaluations using both simulated and real-world recordings confirm the proposed algorithm’s effectiveness, and superiority over baseline methods.