Line-Structured Light Fillet Weld Positioning Method to Overcome Weld Instability Due to High Specular Reflection

Fillet welds of highly reflective materials are common in industrial production. It is a great challenge to accurately locate the fillet materials. Therefore, this paper proposes weld identification and location method that can overcome negative effects high reflectivity. The proposed based on improving semantic segmentation performance DeeplabV3+ network for structural light noise, and, with MobilnetV2, replaces main trunk improve detection efficiency model. To solve problem irregular discontinuous shapes skeleton extracted by traditional methods, an improved closing operation using dilation combined Zhang-suen algorithm was extraction. Then, three-dimensional reconstruction as mathematical model system established obtain coordinates feature points welding-torch angle. Finally, many experiments stainless steel were carried out. experimental results show average errors Y-axis Z-axis 0.3347 mm 0.3135 mm, respectively, error welding torch angle 0.1836° test weld. robust, universal, accurate welds.