室内环境仿人机器人快速视觉定位算法  

An Algorithm for Fast Visual Location with the Humanoid Robot in Indoor Environment

作　　者：吴俊君[1,2] 胡国生[1] 

机构地区：[1]广东食品药品职业学院软件学院,广东广州510520 [2]华南理工大学机械与汽车工程学院,广东广州510641

出　　处：《中山大学学报（自然科学版）》2013年第4期7-13,共7页Acta Scientiarum Naturalium Universitatis Sunyatseni

基　　金：国家自然科学基金资助项目(50975089);高等学校博士学科点专项科研基金资助项目(200805611091)

摘　　要：视觉定位是移动机器人视觉导航的关键问题之一。定位的实时性对视觉导航的性能有较大的影响。在确保定位有效性前提下,提出一种快速的视觉定位方法:采用BRISK(Binary Robust Invariant Scalable KeyPoints)特征作为局部不变性特征点,基于集合理论将场景图像简化为物种种群集合,采用种群相似系数索雷申系数测量场景的相似性,避免繁杂的计算过程,力求简洁快速有效地完成视觉定位。仿真和仿人机器人实验结果表明:在非结构化室内环境下,视觉定位有效性达到99%,场景相似性测量的平均时间0.03 s(每秒33幅图像),验证了该方法的有效性、鲁棒性和良好的实时性。Visual location is one of the most important issues in robot visual navigation. The locating effi- ciency has a large impact on the performance of the visual navigation. A fast visual location method is proposed： BRISK is employed as the local invariant features, the scene model based on aggregate is de- scribed and the similarity of the scenes is measured by Sorensen coefficient to avoid complex computation for efficiency. The method has been verified by simulation and experiment with a humanoid robot in in- door environment. The effectiveness is 99% and the average measuring time is 0.03 s （33 fps）. The re- sult demonstrates that the method has good performance in terms of effectiveness, robustness and real time.

关 键 词：视觉定位 视觉导航 BRISK特征 相似场景 索雷申系数 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]