基于立体视觉的机器人特征跟踪方法  

Research on Autonomous Mobile Robot Feature Tracking Based on Binocular Stereo Vision

作　　者：张晓玲[1] 罗印升[1] 张宝峰[2] 李嘉璐[2] 

机构地区：[1]江苏理工学院,常州江苏213001 [2]天津理工大学,天津300384

出　　处：《激光杂志》2017年第6期116-119,共4页Laser Journal

基　　金：国家自然科学基金项目(61305123)

摘　　要：采用动态跟踪方法实现自主移动机器人的特征跟踪和实时定位,首先利用Harris算子对t时刻与t+1时刻机器人周围场景的左右摄像机序列图像进行特征提取,并对提取出的图像特征点进行初步匹配,最后采用相邻帧间动态比较,获得左右相机前后序列的匹配特征,实现特征跟踪,进行机器人的实时定位。经过实验室内和实验室外环境的验证,该动态跟踪方法实时性较强、精度较高、有较好的定位能力,可以满足机器人的定位需求。The dynamic tracking method is proposed to the feature tracking and location of Autonomous Mobile Ro- bot in the subject. Firstly, Harris operator is used to extract the features of the left and right CCD image sequences at t and t＋l times around the robot. Then, normalized cross-correlation matching is used to preliminarily match the ima- ges feature points, in order to remove the feature points of great differences, and in order to eliminate the erroneously match points. Finally, the method of dynamic comparison between adjacent frames is used to tracking feature in order to realize the real-time location of robot. Experiments show that the dynamic tracking method has good real-time char- acteristic and higher accuracy. It can meet the demand of robot localization.

关 键 词：立体视觉 机器人 特征跟踪 特征提取 匹配 

分 类 号：TN209[电子电信—物理电子学]