计算机与数字工程2018，Vol.46Issue(4)：822-827,6.DOI:10.3969/j.issn.1672-9722.2018.04.039
基于深度卷积神经网络的位置识别方法
BCF:Bags of Convolution Features for Fast Visual Place Recognition
摘要
Abstract
This work proposes a simple visual place recognition pipeline based on a Bag of Convolution Features(BCF), which is implemented by encoding CNN-based features using BoW aggregation.Feature-mapping image is also introduced,which allows directly mapping from regions of the original image to a visual word.Feature-mapping image is then used to perform a fast spa-tial reranking.Furthermore,this system doesn't require any form of task-specified training,all components are generic enough to be used off-the-shelf.The suitability of BCF for visual place recognition is demonstrated,and competitive performance on the challeng-ing Alderley Day/Night dataset and Gardens Point dataset is achieved.关键词
卷积神经网络/词袋模型/特征描述/位置识别Key words
CNN/BoW/feature descriptor/visual place recognition分类
信息技术与安全科学引用本文复制引用
黄于峰,刘建国..基于深度卷积神经网络的位置识别方法[J].计算机与数字工程,2018,46(4):822-827,6.基金项目
高等学校博士学科点专项科研基金(编号:20110142110069)资助. （编号:20110142110069）
