Visual recognition in weld seam tracking based on texture features

The weld seam auto-tracking, which provides important assurance for the automation of welding robots, depends indispensably upon the reliable and real-time vision detection.Considering the seam texture properties, a method was proposed to recognize the weld seam image, which consisted of two steps: image processing by modified sobel edge detector and recognition of seam by applying domain integration.The simulation on Matlab demonstrates that the algorithm developed through this conception can satisfy the requirements of both rapidity and precision.