摘要
基于视觉的机器人自定位方法,相比传统的定位方法具有信息量大、灵活性高、成本低等特点。针对室内移动机器人的自定位,提出了人工地标设计的"四原则",并同时设计了相应的人工地标图形;采用字符编码的方法完成人工地标的定义,选择了基于图像金字塔的最小二乘法形状匹配定位方法;采用"定局-定位-定向"的分层识别策略完成了机器人的自定位。通过实验分析,自定位精度达到亚像素精度,定位方法具有良好的实时性和识别率。
The method for robot self-localization based on vision features heavy information load,high flexibility,and low cost as compared to traditional localization methods. Aimed at self-localization of indoor mobile robots, 'four principles'for artificial landmark design were proposed,while at the same time corresponding artificial landmark graphics were designed. The character encoding method was employed to define artificial landmarks,and the self-localization method of shapes matching of least squares based on image pyramid was chosen. The self-localization of the robot was completed by using the 'foregone conclusion-localization-orientation'hierarchical identification strategy. It is found through an experimental analysis that self-localization reaches Sub-pixel accuracy,and the localization method has good real-time and recognition rate.
出处
《机床与液压》
北大核心
2015年第21期66-69,77,共5页
Machine Tool & Hydraulics
基金
福建省龙岩市重点科技计划项目(2012LY01)
关键词
人工地标
形状匹配
自定位
亚像素
Artificial landmarks
Shape matching
Self-localization
Sub-pixel
