In order to improve the intelligentization level of ship welding surface defect detection and better meet the intelligent manufacturing needs, the intelligent detection system of ship welding surface defect is designed. Through gathering the ship welding surface images and preprocessing the images, combined with the deep transfer learning method, the ship welding surface defect is analyzed according to the defect characteristics, and the detection model is established to conduct the defect target detection for the images, so as to get an efficient means of intelligent detection of ship welding surface defect, which provides driving force for the improvement of shipbuilding intelligentization level and advanced nature.