The invention provides a pipeline weak leakage detection method under a small sample condition and belongs to the technical field of pipeline leakage detection. The method comprises the following steps that: step 1, a real sample set is acquired, and a virtual sample set is generated according to the real sample set; 2, combined feature extraction is performed on the real sample set and the virtual sample set, wherein the combined feature extraction involves seven kinds of statistical features and one group of symbolic transformation features; and 3, a naive Bayesian network pipeline small leakage identification model and a least squares support vector machine pipeline small leakage identification model are established respectively through a naive Bayesian method and a least squares support vector machine method according to the seven kinds of statistical features and the one group of symbolic transformation features, and small leakage detection is carried out on a pipeline through themodels. According to the method of the invention, the weak leakage identification models are constructed from two aspects of increasing the number of weak leakage samples and deeply mining the features of the weak leakage samples, so that the detection accuracy of the weak leakage of the pipeline is greatly improved, and the safe operation of the oil pipeline is guaranteed.