基于L2-Net的回环检测算法
首发时间:2020-05-06
摘要:回环检测算法(Loop closure detection algorithm)是同步定位与建图(simultaneous localization and mapping, SLAM)中的重要算法之一,优化该算法可以提高SLAM系统的性能,对自主机器人的相关研究具有重要价值和意义。本文采用L2-Net直接提取出图像的128维特征向量,去除了邻近帧后,与之前的关键帧的特征向量直接进行相似度计算,省略了词袋模型的使用,完成回环检测算法。实验结果表明,本文的基于L2-Net的回环检测算法与ORB-SLAM2中的回环检测算法相比,在准确性与实时性方面都有较为明显的提升。
For information in English, please click here
Loop closure detection algorithm based on L2-Net
Abstract:Loop closure detection algorithm is one of the important algorithms in simultaneous localization and mapping (SLAM). Optimizing this algorithm can improve the performance of the SLAM system, it has important value and significance for the related research of autonomous robots. This paper used L2-Net to directly extract the 128-dimension feature vector of the image. After removing the neighboring frames, the similarity with the feature vectors of the previous key frames was directly calculated. The loop closure detection algorithm was completed without bag of words model. The experimental results show that the loop closure detection algorithm based on the L2-Net in this paper is improved compared with the loop closure detection algorithm of ORB-SLAM2 in accuracy and real-time performance.
Keywords: Loop closure detection simultaneous localization and mapping L2-Net
引用
No.****
同行评议
勘误表
基于L2-Net的回环检测算法
评论
全部评论0/1000