Robustifying Visual Place Recognition with Semantic Scene Categorization

被引：3

作者：

Arshad, Saba ^{[1
]}

Kim, Gon-Woo ^{[1
]}

机构：

[1] Chungbuk Natl Univ, Dept Control & Robot Engn, Cheongju, South Korea

来源：

2020 IEEE INTERNATIONAL CONFERENCE ON BIG DATA AND SMART COMPUTING (BIGCOMP 2020) | 2020年

基金：

新加坡国家研究基金会;

关键词：

component; visual place recognition; scene recognition; semantic labels; feature descriptors; long term autonomy; visual navigation; hierarchical structure; data segmentation;

D O I：

10.1109/BigComp48618.2020.00-24

中图分类号：

TP301 [理论、方法];

学科分类号：

081202 ;

摘要：

This research proposes an accurate loop closure detection method for the real-time robot navigation in changing light, viewpoint and weather conditions. It focuses on enhancing the accuracy performance of visual place recognition by integrating the semantics of image scene with handcrafted features. This method reduces the computational cost of feature matching process by segmentation of dataset. The results presented depicts that scene recognition can improve the place recognition even in the large viewpoint, weather and light changes

引用

页码：467 / 469

页数：3