Abstract: It is hard to detect weak signals related to natural gas pipeline leakage. In this regard, based on test data derived from simulation of leakages in long-distance natural gas pipelines and with consideration to weak signals submerged in non-Gauss noises, an adaptive detection system for weak signals of pipeline leakage under background of non-Gauss noises is developed. Signals acquired by sensors may contain abundant low-frequency components. By using adaptive filter to track changes in input signals, low-frequency components can be eliminated. In addition, power spectrum of input noise signals may turn more flat. By using matching filters after whitening, weak signals and negative pulse signals submerged in non-Gauss noises can be detected. Results obtained in two simulation tests for pipeline leakage show that this new algorithm can effectively detect weak signals of pipeline leakage under background of non-Gauss noises.