基于显著路标与局部描述子的机器人视觉导航  

Robotic visual navigation based on salient landmarks and local descriptor

作　　者：刘冬[1] 丛明[1] 高森[1] 杜宇[1] 

机构地区：[1]大连理工大学机械工程学院,辽宁大连116024

出　　处：《华中科技大学学报（自然科学版）》2013年第S1期258-261,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(50675027)

摘　　要：提出一种基于生物学视觉注意机制与图像局部特征技术的机器人视觉导航算法.该算法通过建立基于状态的环境拓扑地图进行路径学习,计算输入场景的显著图获得对环境变化相对稳定的自然路标,提取其局部描述子特征,基于描述子匹配对的横向坐标进行机器人定位并预测机器人的行为模式,实现路径的跟踪导航.实验表明:该算法无须进行视觉传感器校正,能够较好地完成机器人跟踪导航任务,并且对环境中的动态信息具有一定的鲁棒性.A robotic visual navigation algorithm was proposed based on biologically-inspired visual attention selection mechanisms and image local features representation.This algorithm performed route learning by establishing state-based topological map,and computed saliency map of the input scene to obtain stable natural landmarks.After that,by extracting local descriptors,the algorithm utilized the horizontal coordinate of the matched descriptors pairs to process robot localization and predict the robot's action,and realized route tracking and navigation.The experimental results show that the proposed algorithm does not need camera calibration,and can perform good tracking and navigation task of the mobile robot,and has robustness for dynamic information of environment.

关 键 词：移动机器人 视觉注意 局部描述子 导航 算法 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]