Automatic gripping strategy for package yarn end by composite robot based on RGB-D camera

Abstract

In the pretreatment process of warping, it is extremely tedious work to take out a large number of package yarn ends from the warping frame and pull them to the warping machine; this task is currently completed manually. To automate this process, in this work, a composite robot equipped with a red-green-blue-depth (RGB-D) camera is adopted to automatically identify, locate, and grasp the package yarn ends on the warping frame. This work builds on previous research, where the yarn end was made to protrude from the package surface through the application of negative pressure. With the aim of addressing the challenge of determining depth accurately, owing to the weak features of single yarn ends, a novel approach is proposed that combines point cloud and image processing to estimate the precise spatial pose of the robot when grasping yarn ends. First, point cloud processing enables precise identification and spatial localization of the circular contour on the package end. Subsequently, image processing, with the introduction of a reinforcement operator for weak features, facilitates extraction of the yarn end and accurate determination of its relative position with the center of the circular contour on the package end. Finally, an automated grasping path for the composite robot is planned, based on geometric information about the spatial circle on the package end and the relative positioning between it and the yarn end. The accuracy of identifying and localizing yarn ends detached from the packages, as well as the effectiveness of the automatic robot gripping strategy, are validated through simulation and experimentation.