Abstract: In view of shortcomings such as low detection efficiency, difficulty in realizing automation and intellectualization in the process of evaluating weld defects of thrust chamber by conventional film photography, an automatic defect detection method based on X-ray digital imaging detection technology was proposed, and a DR digital imaging detection system was built. Improved Faster R-CNN network was used to establish a digital image weld defect recognition model, recognition accuracy of the model was tested and the model was deployed on the DR detection system. The research results showed that the trained improved Faster R-CNN network model could accurately identify four typical brazing seam defects in DR images, recognition accuracy reached more than 93%, and defect recognition time of a single image did not exceed 2 s. Improved Faster R-CNN network model was used to perform computerized intelligent evaluation of DR digital imaging inspection photos of brazed joints in liquid rocket motor inference chambers, and it was demonstrated that the model could be deployed in DR inspection systems to achieve intelligent online detection of defects.