Abstract: The extraction process of molten pool signal and the method of removing interference signal to improve detection stability are studied in online quality detection of laser scanning welding technology. A clear image of molten pool is obtained by constructing a coaxial vision detection system with Nd: YAG laser as the heat source. For the interference of many factors on quality monitoring in production process, a method of filtering clutter and extracting key quality information is proposed by using related digital image technology. The threshold value is set reasonably by the method of maximum inter-class variance, and the gray scale image is converted into a binary image. The enhancement operator and maximum interconnection region method are used to remove the same gray level noise points caused by welding spatter effectively and obtain the accurate molten pool boundary. The influence of welding fixture and splash shape on boundary identification is analyzed, and reasonable working condition standard is set to improve the stability of quality control and reduce false alarm rate

Key words: laser scanning welding, molten pool, spatter, boundary extraction, quality detection