High-accuracy 3D image stitching for robot-based inspection systems

Abstract

Robot-based inspection systems are flexible and fast, but lack the required accuracy due to the typically utilized industrial robots. While this contradiction is most often dealt with calibration of the robot or tracking of the tool, the presented approach will use neither. Instead, the gathered point clouds of the 3D sensor are stitched together in order to give a local reference in regard to the starting point of the robot's movement. For the application of inspecting sheet metal parts, plain and featureless component surfaces must be taken into account. Therefore a texture projection is applied to provide additional information to the stitching algorithm. The resulting procedure can be used to calculate the sensor's movement with high accuracy without the need to consider the used robot.