Human-assisted rapid modeling of welding environment

Providing 3D models of welding environment to robot is a critical component in remote welding. This paper introduces U man-machine interaction modeling system based on stereo vision. To avoid problems caused by untextured surface of welding environment, we project light patterns onto the welding work piece to provide feature, points for automatic stereo matching. When stereo image pair of the environment are acquired, automatic point extraction and matching is applied For planar work piece, the operator segment point data to isolate points on the plane from non-planar points. In the process of planar surface fitting, we adopts least squares technique and Principal Component Analysis (PCA) is used for cylinder surfaces fitting. The result shows that the maximum fitting deviation of planar surface is less than 3mm. The deviation of estimated cylinder radius is 0. 7mm.