Abstract: With the increasing demand for high-quality thick plate welding in high-end equipment manufacturing, narrow-gap welding has garnered considerable attention due to its high efficiency and superior weld quality. However, the deep and narrow groove geometry, coupled with the complex welding environment, presents significant challenges particularly in ensuring reliable sidewall fusion. A comprehensive review of feature extraction and recognition techniques in narrow-gap welding image processing was provided, with a focus on recent advancements and persistent technical challenges in visual sensing for welding applications. Specifically, the fundamental differences between active and passive visual sensing methods were examined. Their respective advantages and roles in weld process monitoring were evaluated. Furthermore, the key image processing techniques in narrow gap welding was discussed, including image preprocessing, segmentation, and edge detection, while critically analyzing the limitations of conventional algorithms in complex welding environments. Finally, in response to the existing challenges, the potential future directions for feature extraction and recognition in narrow gap welding were explored, offering valuable insights for advancing intelligent welding technologies.