Efficient Visual Place Recognition by Adaptive CNN Landmark Matching
Efficient Visual Place Recognition by Adaptive CNN Landmark Matching
复制标题
DOI：
10.3837/tiis.2021.11.012
复制
发表时间：
2021-11
期刊：
影响因子：
--
通讯作者：
Yutian Chen;Wenyan Gan;Yi Zhu;Hui Tian;Cong Wang;Wen-Jun Ma;Yunbo Li;D. Wang;Jixian He
中科院分区：
文献类型：
--
作者：
Yutian Chen;Wenyan Gan;Yi Zhu;Hui Tian;Cong Wang;Wen-Jun Ma;Yunbo Li;D. Wang;Jixian He
视觉位置识别（VPR）是移动的机器人导航与定位的一个基本而又具有挑战性的任务。现有的VPR方法通常是基于图像描述子的某种成对相似性，因此它们对视觉外观变化敏感，计算量也很大。本文提出了一种简单而有效的四步方法，实现了自适应卷积神经网络（CNN）的VPR地标匹配。首先，基于从现有CNN模型中提取的特征，选择具有较高显著性分数的区域作为地标。然后，根据潜在地标的坐标位置，通过去除不匹配的地标对来改进地标匹配
Visual place recognition (VPR) is a fundamental yet challenging task of mobile robot navigation and localization. The existing VPR methods are usually based on some pairwise similarity of image descriptors, so they are sensitive to visual appearance change and also computationally expensive. This paper proposes a simple yet effective four-step method that achieves adaptive convolutional neural network (CNN) landmark matching for VPR. First, based on the features extracted from existing CNN models, the regions with higher significance scores are selected as landmarks. Then, according to the coordinate positions of potential landmarks, landmark matching is improved by removing mismatched landmark pairs