服务机器人目标同时识别与位姿判定研究  

Object Simultaneous Recognition and Pose Estimation for Service Robot

作　　者：吴培良[1] 孔令富[1] 李海涛[1] 

机构地区：[1]燕山大学信息科学与工程学院,河北秦皇岛066004

出　　处：《工程图学学报》2010年第5期81-88,共8页Journal of Engineering Graphics

基　　金：国家自然科学基金资助项目(60975062);国家"863"计划CIMS主题资助项目(2006AA04Z212)

摘　　要：目前机器人尚难以自主建立起目标姿态的概念,通过人机交互方式辅助机器人对目标建模则不失为一种可行方案。提出了一种基于新型目标模型的离线/在线两步目标识别与位姿判定系统。离线阶段采用人机交互方式构建目标模型,在线阶段通过场景图像与目标模型的特征点对应即可自动发现目标,并计算其相对于机器人的位姿。实验证明,该系统能够有效完成对纹理目标的同时识别与位姿判定。At present it is still difficult for service robot to estimate the pose of different objects autonomously,and guiding service robot to build the object model by human-machine cooperation techniques seems to be a feasible scheme.A novel object model is proposed and based on which a two-step of offline/online system is presented to recognize the object and estimate its pose simultaneously.At the offline step the object model is built guided by human-machine cooperation techniques,at the online step the object is detected and the pose is estimated according to the correspondence between the object model and a captured image.The experiments show that the proposed system can recognize the textural object and estimate its pose simultaneously and effectively.

关 键 词：计算机应用 新型目标模型 SIFT特征 目标同时识别与位姿判定 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]