Human machine interface in weld seam detection using an RGB-D camera

The set-up time in robotic welding needs to be reduced so that it can be used for limited production runs. One challenge in achieving this reduction is the automatic localisation of the weld seam. A method is developed in this paper to locate the weld seams by human-machine interface. In particular, this method uses an RGB-Depth sensor, such as the Microsoft XBOX Kinect sensor, to identify a human hand that is pointing at the weld seam. A 'hybrid hand detector' is developed to recognise the human hand and fingers using both the depth and colour information. In turn it enables the identification of the weld seam. Experimental results shown that the weld seam can be reliably identified using the proposed method at near real time speeds.