项目名称: 自然场景下机器人大范围视觉伺服研究
项目编号: No.61203345
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 自动化学科
项目作者: 辛菁
作者单位: 西安理工大学
项目金额: 24万元
中文摘要: 目前在自然场景下大范围机器人视觉伺服研究方面存在着需要人工标记,大范围伺服过程中目标易偏离摄像机视场等问题。为此,本项目提出基于仿射不变特征的机器人自适应免疫变焦视觉伺服控制策略。利用仿射不变特征检测算法检测出适合自然场景下机器人大范围视觉伺服任务的仿射不变特征,在此基础上结合改进的极线几何约束设计鲁棒的特征匹配算法、实现自然场景下无人工标记的一般物体的特征提取。通过在机器人视觉伺服技术中引入主动焦距控制,动态调整摄像机视场的大小以保证特征点始终在视场内同时在伺服末期获得较高的目标局部分辨率。利用投影不变性构造一个与摄像机内参数无关只与摄像机与目标的相对位置相关的误差函数,基于李雅普洛夫稳定性理论自适应估计出变焦图象雅可比矩阵,在此基础上设计出基于仿射不变特征的机器人自适应免疫变焦视觉伺服控制律,实现自然场景下机器人大范围视觉伺服。本项研究不仅具有重要的理论创新意义而且具有广阔的应用前景。
中文关键词: 自然场景;大范围视觉伺服;仿射不变特征;变焦控制;不变空间
英文摘要: At present, in the most robot visual servoing researches, object feature extraction process is oversimplified by using a fiducial markers and object could get out of the field of view of the camera during the robot large displacement motion. In order to solve the above problems, a robot adaptive immune zooming visual servoing method using affine invariant image features is proposed. Firstly, in order to deal with large viewpoint changes and that do not rely on specific markers in various illumination conditions and achieve feature extraction for the more general object in the nature scenes , affine invariant features, which can be suitable to the task of robot large displacement visual servoing under the natural scenes, are detected by the affine invariant feature detector, and feature matching is achieved by a robust image matching algorithm based on the improved epiploar geometry constraints. Secondly, a zoom control mechanism is introduced into the robot visual servoing system, which enable dynamic adjustment of field of view of the camera to keep all the feature points of the object in the field of view of the camera and get the high object local resolution in the end of visual servoing. Finally, system error function can be constructed by projective invariant property, this error is invariant to the change
英文关键词: nature scenes;large displacement visual servoing;affine invariant feature;zoom control;invariant space