Development of an Intelligent Weld Surface Defect Detection System: Design and Application

Early evaluation through the detection of surface defects in welds is essential to ensure the quality of weldments. In this study, considering the complexity of weld surface inspection, we designed a high-performance inspection system that combines image processing, defect detection, and model evolution functions. This system facilitates the engineering application of image super-resolution reconstruction, defect detection, and semi-supervised learning methods in real-world weld detection scenarios through the integrated weld surface inspection module and a visually interactive client. The experiment results demonstrate that the system can perform real-time processing, with a speed surpassing 30 frames per second, for super-resolution and detection evaluation of input images. Furthermore, it has autonomous learning capabilities, which continuously improve detection precision and reduce the miss rate during long-time detection.