A Novel Data Association Approach for Slam in Dense Features Environment

摘要：

　　The correct data association increases robustness and enhances accuracy in SLAM.It attracts much attention over the past decades.In this paper,an efficient data association approach for SLAM based on JCBB and K-means clustering is presented.Firstly,in order to decrease the computation complexity and preserve high matching accuracy in dense feature environment,the small correlative measurements are separated into several groups by K-means clustering.The number of groups is depended on the characteristics of the environment.Secondly,each group local correlations are produced by JCBB and ICNN,respectively.Finally,all local correlations are put together to find the most joint compatibility one as the global correlation.The experimental results demonstrate that the proposed approach can achieve high matching accuracy and low computation time than the existing state-of-the-art methods.

关键词： SLAM data association K-means clustering JCBB ICNN

作者: Yao Cong Du Jianyu Zhao Defang Wang Ran

作者单位: China FAW Group Corporation R&D Center

会议类型: 国内会议

会议名称: 第19届亚太汽车工程年会暨2017中国汽车工程学会年会

会议地点: 上海

会议语种:英文

页码: 11-16

在线出版日期: 2017-10-24（万方平台首次上网日期，不代表论文的发表时间）