Features for image processing of OCT images for seam tracking applications in laser welding

Abstract

The coaxial integration of optical coherence tomography (OCT) enables the determination of surface topography and measurement of object features along the optical path within laser machining. The measurement of surface information from the processed workpiece allows for the identification of features from the joint configuration and the subsequent control of the welding process by seam tracking. State-of-the-art seam tracking approaches are based on monochromatic cameras or laser triangulation. Typically, these approaches apply line segmentations for the identification of the joint position. The interferometric measurement method of OCT gives rise to the identification of new features for image processing in seam tracking. In this work, we identify specific noise components and features based on the theoretical background of OCT for image processing in seam tracking applications. Two different features are derived for the detection of arbitrary joint configurations with corresponding systematic image processing approaches. In the first step, we show the applicability of line detection methods for feature detection of arbitrary joint configurations. The necessary evaluation algorithm for case sensitivity and limitations (e.g., chamfer) in detecting different joint geometries are discussed. In the second step, we show an approach in feature extraction with feature detectors (e.g., ORB, SURF) for a new image feature. Here, significant image space from (multiple) reflections at the joint position is used for joint detection. The detectability is discussed depending on the joint configuration. The results show good suitability of both features for seam tracking applications.