Abstract: To address the problems of low efficiency and excessive reliance on imaging in traditional non-destructive testing technologies for defect recognition of polyethylene (PE) pipe fusion joints, an intelligent detection method integrating microwave line-scan data and a hierarchical recognition machine learning architecture was proposed. The S₁₁ reflection coefficients of the pipe joints were acquired through a custom-built microwave detection device, and the multi-dimensional features of the signals in the time domain, frequency domain, and time-frequency domain were extracted. The MI-RFECV algorithm was adopted to dynamically select the key features, and a hierarchical classification model was constructed to identify different types of defects. The performance of the model was validated on an independent test set and compared with the results of microwave imaging and tensile tests. The results indicate that the method exhibits excellent performance in defect recognition, with an accuracy A of 98.6%, a recall R of 98.8%, and an F1-score of 98.2%, and it possesses good generalization ability in the detection of various defects. This research effectively improves the intelligence level of defect detection for PE pipe joints, reduces the reliance on imaging and manual judgment, and provides a new idea for the automatic quality assessment and operational safety assurance of PE pipeline systems.