Abstract: When a compact underwater platform carries a passive acoustic payload for detecting non-cooperative signals, platform-related interference often forms a set of narrowband spectral lines within the operating band, comprising both stationary and nonstationary components. These components may exhibit slow drifts in frequency and amplitude with changing operating conditions, which degrades the salience of weak tonal components, impairs near-frequency parameter estimation, and complicates false-alarm control. To address these challenges, this paper proposes a space–frequency joint detection framework tailored to platform-interference backgrounds. In the frequency domain, an adaptive interference representation is constructed by exploiting the quasi-harmonic consistency and clustered structure of interference line sets, and line-level suppression is applied to reduce masking of candidate tonal components. In the spatial domain, a weakly frequency-coupled discriminant is designed by leveraging differences in spatial coherence between the desired signal and near-field platform interference across array observations, thereby improving resolvability under co-frequency or near-frequency conditions. Simulation and experimental results demonstrate that the proposed space–frequency joint strategy remains robust in the presence of platform interference with fluctuating statistical characteristics, and it provides a generalized approach for real-time detection in practical engineering scenarios.