基于人工路标的鲁棒的室内机器人定位方法  

Robust method of indoor robot localization based on artificial landmark

作　　者：朱齐丹[1] 刘鹏[1] 蔡成涛[1] 

机构地区：[1]哈尔滨工程大学自动化学院,哈尔滨150001

出　　处：《计算机应用》2017年第A01期126-130,136,共6页journal of Computer Applications

基　　金：国家自然科学基金资助项目(61175089;61203255)

摘　　要：针对移动机器人室内定位问题,提出了一种对光照和环境变化具有良好鲁棒性的快速定位方法。使用全景视觉系统获得环境信息,首先将特征匹配算法应用到人工路标的识别中,使路标识别不受光照和环境变化影响,且针对该算法设计了利于识别和定位的路标;然后利用设置动态检测窗口的方法缩小了路标检测区域,解决了全景图像的特征提取运算量大的问题,提高了算法识别率;最后采用路标模板实时更新的策略,消除了视角变化对识别率的影响。在复杂的室内进行了多个场景的对比实验,结果证明本路标识别算法相比颜色识别算法具有更高的鲁棒性,且定位的实时性与精度能够满足室内机器人的定位需要。According to the indoor localization of mobile robot, a rapid localization method was proposed, which has good robustness to illumination and environmental changes. an omni-directional vision system was used to obtain environmental information. Firstly, the feature matching algorithm was applied to the landmark recognition in order to make the recognition not be affected by illumination and environmental changes. Then the landmark was designed which is conducive to recognition and localization according to this algorithm. Secondly, dynamic detection window was set to reduce the landmark detection area and solve the problem caused by the large amount of computation, the recognition rate was improved. Finally the affection of the angle of view on the recognition rate was eliminated by updating templet in real-time. In complex indoor environment,many experiments under different scenarios indicate that this method is more robust than color recognition, and its real-time performance and high accuracy of positioning can meet the needs of indoor robot positioning.

关 键 词：移动机器人 室内定位 全景视觉 特征匹配 路标识别 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]