Development of groove recognition algorithm with visual sensor. Practical development of visual sensor for welding robot (1st Report)

Abstract

Summary To improve the automation or artificial intelligence of welding operations, it is necessary to develop high-performance sensors. Laser visual sensors are designed to recognise the configuration or detect the position of a weld by irradiating a slit light on to the welded joint, receiving the light with a CCD camera, and converting the two-dimensional information into three-dimensional information by image processing. The visual sensors for welding robots must be free from the influences of welding process disturbances, such as arc light, heat, spatter and fume. This paper describes optical component arrangements and image processing algorithms. The optical component of a sensor head helps avoid optical window contamination caused by spatter or fume. The image processor used can process image data quickly and accurately even if welding arc light illuminates the workpiece surface. The accuracy of weld position detection is within ±0.2 mm.