A Novel Sub-pixel Refinement Method for Fillet Weld Under Structured Light Vision
Abstract Intelligent welding robots based on structured light vision are widely used in industrial production. With the demands of low cost, miniaturization and flexibility, the development of embedded structured light vision systems for seam tracking is a general trend. The core is how to efficiently and precisely position weld seam with low-configuration hardware. Sub-pixel refinement can break through the limitation of physical resolution, while also reducing hardware cost to achieve the required accuracy. To fill the gap in the sub-pixel refinement for fillet weld joint, a novel sub-pixel refinement method for fillet weld joint under structured light vision is proposed, which can sub-pixel refine the fillet weld joint under various working conditions. The main novelties of the proposed method include: (1) a novel sub-pixel refinement method for fillet weld joint by using Mean Shift, weighted least square and directional maximum projection is proposed, which is robust, universal, and accurate. (2) A directional maximum projection algorithm for refining weld is proposed for the first time. (3) The method can accurately refine fillet weld joint with low-resolution image. The proposed method is robust, universal, and accurate, and as demonstrated by the following performances: the average and maximum bias are 0.73 and 3 pixels in the accurate test, positioning accuracy rate is 100% in the test of noise-free, rusty, highly reflective and arc radiation-and-spatter working conditions. the method can be expanded to a common sub-pixel refinement method for structured light intersections through simple transformation.
