| 基于深度学习的视觉SLAM回环检测方法 |
| |
| 引用本文： | 余宇,胡峰. 基于深度学习的视觉SLAM回环检测方法[J]. 计算机工程与设计, 2020, 41(2): 529-536 |
| |
| 作者姓名： | 余宇 胡峰 |
| |
| 作者单位： | 重庆邮电大学 计算机科学与技术学院,重庆 400065;重庆邮电大学 计算机科学与技术学院,重庆 400065 |
| |
| 基金项目： | 重点产业共性关键技术创新专项基金项目;国家自然科学基金;重庆市基础与前沿研究计划;国家重点研发计划 |
| |
| 摘 要： | 现今主要的视觉SLAM回环检测方法是基于人工标记特征点算法进行图像间匹配,在复杂环境下会出现准确率急速下降的问题。针对此问题,结合卷积神经网络和局部敏感哈希算法,提出一种基于深度学习的回环检测方法。基于回环检测中的图像相似性判断策略构建图像特征向量集,运用级联的余弦距离哈希函数进行回环检测。实验结果表明,该方法较传统方法有着更高的准确率与速率,更好满足了视觉SLAM系统对消除累计误差和实时性的要求。
|
| 关 键 词： | 同步定位与建图 回环检测 卷积神经网络 局部敏感哈希 图像特征提取 |
Loop closure detection method based on deep learning for visual SLAM |
| |
| YU Yu,HU Feng. Loop closure detection method based on deep learning for visual SLAM[J]. Computer Engineering and Design, 2020, 41(2): 529-536 |
| |
| Authors: | YU Yu HU Feng |
| |
| Affiliation: | (School of Computer Science and Technology,Chongqing University of Posts and Telecommunications,Chongqing 400065,China) |
| |
| Abstract: | The common ways for loop closure detection have low accuracy in complex environments.To solve this problem,a method based on deep learning was proposed.Convolutional neural network and locality sensitive hashing algorithm were used for image feature vector set construction and loop closure detecting respectively.Experimental results show that the proposed method achieves higher accuracy and speed than traditional methods. |
| |
| Keywords: | simultaneous localization and mapping loop closure detection convolutional neural network locality sensitive hashing image feature extraction |
| 本文献已被 维普 万方数据 等数据库收录！ |
