论文部分内容阅读
为提高机器人的视觉感知能力,提出了一种新颖的基于物体逻辑状态推理的未知物体视觉分割方法.在语义层定义物体逻辑状态空间,根据机器人抓取动作反馈对物体逻辑状态进行推理.在数据层利用RGB-D摄像头采集生成场景3维着色点云,基于物体放置于支撑平面的假设,对所有可能的物体点进行空间聚类和分割,得到初始未知物体集并生成物体初始逻辑状态表示.当物体逻辑状态发生变化时,根据设定规则结合点集运算对物体点云进行重新分割,物体点云集的变化又用于指导物体逻辑状态空间的更新.在7自由度移动机械手系统上,进行了真实积木模拟环境下未知物体的视觉分割与抓取实验,实验结果证实本文方法可以有效提升机器人的视觉感知能力.
In order to improve the robot’s visual perception ability, a novel visual object segmentation method based on logical state reasoning is proposed, in which the logical state space of the object is defined in the semantic layer, and the logical state of the object is inferred according to the feedback of robotic crawling.In the data Based on the assumption that the object is placed on the support plane, all the possible object points are spatially clustered and segmented by using RGB-D camera acquisition to generate the initial three-dimensional color point cloud and obtain the initial unknown object set and generate the initial logic state representation of the object. When the logical state of the object changes, the object point cloud is re-divided according to the set rule combined with the point set operation, and the change of the object point cloud set is used to guide the updating of the logical state space of the object. On the 7-DOF mobile robot system, The experiment of visual segmentation and grasping of unknown objects under the real building block simulation environment shows that the proposed method can effectively enhance the visual perception of the robot.