A Slope K Method for Image Based Localization

摘要：

In this paper, we present a SIFT based Slope K method which is faster and more robust than the classical SIFT in landmark based localization. First, the slope k value can be used to erase mismatched feature points (outliers) of the two compared images. Second, the y position is determined by the slope k value. Therefore, the Slope K method is able to localizes about twice as more accurate as the classical SIFT. Another advantage of the proposed method is that the number of database images needed to be matched is significantly reduced, compared to the classical SIFT. Therefore the time cost is approximate 4 times less than that of the classical SIFT.

作者: Hong Liu Haitao Yu Yuexian Zou Zhenhua Huo

作者单位: Key Laboratory of Machine Perception and Intelligence and the Key Laboratory of Integrated Micro-sys Key Laboratory of Integrated Micro-system,Shenzhen Graduate School,Peking University,Shenzhen 518055

会议类型: 国际会议

会议名称: 2009 IEEE International Conference on Robotics and Biomimetics(2009 IEEE 机器人与仿生技术国际会议 ROBIO 2009)

会议地点: 桂林

会议语种:英文

页码: 535-538

在线出版日期: 2009-12-19（万方平台首次上网日期，不代表论文的发表时间）