In underwater optical wireless communication （UOWC） systems， the low signal-to-noise ratio causes received optical signals to be submerged in strong noise. To solve this problem， a detection method for underwater weak optical signals based on adaptive stochastic resonance （ASR） is proposed. In this study， the characteristics of underwater weak light signals and the detection mechanism of stochastic resonance are analyzed. The concept of twice sampling is introduced to enable the application of stochastic resonance to optical signal detection at any frequency. Additionally， the parameters that affected the stochastic resonance system are evaluated， and a detection method combining the improved ant colony algorithm and stochastic resonance is proposed. The system dynamically adjusts the system parameters based on the detection performance， such that the entire system achieves the optimal matching state to realize ASR. To verify the effectiveness of the method， a UOWC experimental system based on an avalanche photodiode is built. The experimental results show that the bit error rate is 5×10^-4 when SNR=-4.559 5 dB. Thus， underwater weak light signal detection based on ASR is proved to significantly improve the bit error rate performance of UOWC receivers.