|
基于三维点云聚类边缘点的回环检测算法 |
Loop detection algorithm based on 3D point cloud clustering for edge points |
|
DOI: |
中文关键词: 回环检测 SLAM 三维点云 描述符 词袋 聚类 边缘点 |
英文关键词: loop detection SLAM 3D point cloud descriptor word bag clustering edge point |
基金项目: |
|
摘要点击次数: 408 |
全文下载次数: 298 |
中文摘要: |
回环检测是机器人同步定位与建图(SLAM)的重要组成部分,可以消除 SLAM过程中产生的累积误差。在激光 SLAM领域,传统的回环检测方法可能无法有效地实时识别回环,并且无法校正完整的六自由度环路姿态。为此提出一种基于三维点云聚类边缘点的回环检测算法。首先利用三维点云的边缘点进行聚类生成描述符,随后将描述符以单词的形式存储到词袋,采用哈希表构建单词与位置的一对一直接关联,最后通过逆向索引进行位置识别。该方法不仅能有效地识别重访的环位,而且能实时校正整个六自由度环路姿态。分别在 M2DGR数据集、 KITTI数据集和真实环境中进行了实验,结果表明本文算法具有旋转不变性和更高的准确率。 |
英文摘要: |
Loop detection is an important part of synchronous localization and mapping(SLAM),which can eliminate the accumulated errors during SLAM process. In the field of laser SLAM,traditional loop detection methods may not be able to effectively identify loops in real-time and correct the complete 6-DOF loop attitude. Therefore,a loop detection algorithm based on 3D point cloud clustering for edge points is presented. Firstly,the edge points of 3D point cloud were clustered to generate descriptors which were then stored in the word bag in the form of words,and the one-to-one direct association between words and positions was constructed by using Hash table. Subsequently,the position recognition was carried out via reverse indexing.This method can not only detect the revisited loops effectively but also correct the position and posture of the whole 6-DOF loop in real-time. The results of experiments on M2DGR and KITTI datasets as well as in real environment indicate that the proposed algorithm has rotational invariance and higher accuracy. |
查看全文
查看/发表评论 下载PDF阅读器 |
关闭 |