闭环检测中词袋与词对袋的对比研究  

A comparative research on BoW and BoWP in loop closure detection

作　　者：曾桂萍 孙作雷[1] 潘盼 

机构地区：[1]上海海事大学信息工程学院,上海201306

出　　处：《微型机与应用》2017年第17期21-23,30,共4页Microcomputer & Its Applications

摘　　要：从词袋技术(Bo W)入手探究拓扑地图中的闭环检测算法,在性能和时间效率两个因素系统地对比了词袋技术与词对袋技术(Bo WP),词袋技术是以Dorian Gálvez López提出的DLoop Detector算法为参考,词对袋(Bo WP)技术是以Nishant Kejriwal提出的高性能闭环检测为原始算法,主要凸显了词对袋技术在性能上的优越性。词袋技术的优点是技术成熟易实现,但存在感知混叠问题。通过额外创建一个由词对组成的词典来克服传统词袋法中因矢量的量化导致的感知混淆的限制。实验表明,词对袋方法比Dloop能提供更好的召回性能,并且减少了算法的计算时间及复杂度。The paper explores loop closure detection algorithm in the topological map from Bag of Words(BoW) technology,and systematically compares the performance in time and efficiency between Bag-of-Word and Bag of Word Pairs(BoWP). The BoW technology reference the DLoop Detector algorithm proposed by Dorian Gálvez López,because it's mature technology is easy to realize. The BoWP technology is based on a high performance loop closure detection algorithm,which proposed by Nishant Kejriwal. It highlights the superiority in the BoWP technology on the performance,by creating an extra dictionary of word pairs to overcome the limitation of perceptual aliasing due to vector quantization in the traditional bag of word method. The experimental results show that the BoWP can provide better recall performance than the typical Dloop,besides,it can reduce the computational time and complexity of the algorithm.

关 键 词：词袋 词对袋 闭环检 同步定位和构图 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]