As an important infrastructure for energy transportation, pipelines play a crucial role in ensuring safety and promoting economic development. However, as energy transportation systems remain in service for long periods, various defects may gradually emerge, leading to economic losses or even environmental damage. Recently, object detection methods based on deep learning have achieved promising results in object detection fields. However, due to the lack of mechanistic characteristics and insufficient ability to extract defect features under different operating conditions, deep learning-based defect detection methods still struggle to achieve satisfactory detection results in the defect detection field. To address these issues, this paper proposes an intelligent defect detection method based on the magnetic flux leakage (MFL) mechanism, which effectively combines expert knowledge to enhance the model's detection accuracy. First, a classification regression decoupling (CRD) module is designed to enhance the model’s ability to capture discriminative features and spatial information according to the characteristics of different tasks, thereby maintaining high classification accuracy with precise regression positioning. Second, a defect physical characteristics (DPC) loss is designed to effectively guide the model in learning key features related to defect physical characteristics, thereby further improving detection performance.