Abstract: ‍ ‍With the rapid development of artificial intelligence and other fields， unmanned aerial vehicles play an increasingly important role in various fields， especially modern warfare. With their strong flexibility and high concealment advantages， UAVs have a subversive impact on modern operations. For the UAV to carry out battlefield reconnaissance， surveillance， communication， electronic jamming， radar decoy， laser guidance， and other tasks， timely detection of enemy radiation signals is particularly important. The stochastic resonance method is one of the commonly used weak signal detection methods， which can make the input signal and noise produce a synergistic effect under certain conditions by transferring the energy of noise to the signal. The technique of stochastic resonance detection for narrow-band signals has matured. However， for weak broadband communication signals in scenarios such as surveillance and communication of small-load UAVs， the existing stochastic resonance detection methods directly use the time-domain energy detection method of narrow-band signals and do not design a special detection method based on the characteristics of broadband signals， so the detection performance is limited under a low signal-to-noise ratio. This study leveraged commonly used broadband communication signals as the research object to meet the detection requirements of weak broadband communication signals. A stochastic resonance signal detection method based on the local spectrum of broadband signals was proposed for single-channel receiving scenarios. Compared to traditional stochastic resonance signal detection methods， the performance of this method was improved by about 6 dB. Secondly， with lower SNR， a two-channel cooperative detection method combining stochastic resonance and cross-correlation detection was proposed， and the system parameter optimization criteria based on the maximization of the mutual relationship between the two channels after stochastic resonance was proposed. Simulation analysis in typical scenarios shows that the local spectrum method can effectively improve the detection performance of weak broadband communication signals with low SNR received by a single channel. For broadband communication signals with lower SNR， dual-channel cross-correlation detection using stochastic resonance processing can further improve detection performance.