Abstract: In order to improve the accuracy of the defect quantification technology for pipeline magnetic flux leakage internal detection, a series of feature extraction methods for pipeline axial, radial and circumferential directions was designed based on the defect magnetic flux data collected by the three-axis magnetic flux leakage internal detector, providing a data basis for the high-precision quantification of subsequent defects. Aiming at the problem of feature redundancy under the task of quantifying defects with different sizes, this paper proposes a feature selection method based on nearest neighbor component analysis, which can effectively eliminate irrelevant features in the original feature set. In the quantization of magnetic flux leakage defects based on support vector regression, this paper combines the improved bat algorithm to optimize the parameters of the support vector regression. The results show that the designed quantization method can effectively reduce time complexity and improve the accuracy of defect quantification to a certain extent.