During the "14th Five-Year Plan" period, the Ministry of Industry and Information Technology of the State put forward the industrial Internet development plan to accelerate the integration of industrial production with 5G, big data, artificial intelligence and other information technologies. In the field of weld defect detection, it is still a low-efficiency specialized personnel detection. The main detection method. With the development of China's industrial manufacturing industry, the demand for weld defect detection in shipbuilding, power generation, aerospace, boiler and other industries has increased dramatically, and the emergence of intelligent weld defect detection technology is imminent. This work-an intelligent detection system for industrial weld defects based on X-ray images, combined with in-depth learning and traditional computer vision methods, has designed a set of schemes to detect defects in weld X-ray images generated by X-ray inspection technology. The aim is to transform human observation of defects in X-ray images into computer detection and identification of weld defects in X-ray images, and present the detection results on the original X-ray images for further review by technicians. At present, there are 12 kinds of internal defects in welds, such as bubbles, weld bumps, cracks, etc. Different types of defects have different physical properties and morphological texture characteristics. This work designs a universal detection method for various defects based on deep learning, integrates the recall rate, accuracy rate, time and other factors of the detection algorithm, and adds auxiliary detection methods based on traditional visual algorithms to form the final scheme. Finally, after experiments, the detection scheme of this work still shows strong robustness in the presence of noise, blur and low contrast problems in the input X-ray image. As an ideal intelligent system for weld defect detection, this product can not only save additional manual detection costs, but also is a technological innovation in the weld detection process, which has broad application prospects and can produce significant economic and social benefits.