人工智能培训

搜索

论文代码开源:松耦合半直接单目SLAM(Loosely-Coupled Semi-Direct Monocular SLAM)

[复制链接]
admin 发表于 2018-7-28 12:25:50 | 显示全部楼层 |阅读模式
admin 2018-7-28 12:25:50 2207 0 显示全部楼层
人工智能论文代码开源:松耦合半直接单目SLAM(Loosely-Coupled Semi-Direct Monocular SLAM)请注意该人工智能论文代码开源在github,大部分是python写的,框架可能是tensorflow或者pytorch。我们提出了一种新颖的半直接方法,用于单目同时定位和映射(SLAM),它结合了直接和基于特征的方法的互补优势。所提出的管道松散地耦合直接测量和基于特征的SLAM以执行三个级别的并行优化:(1)光度束调整(BA),其联合优化局部结构和运动,(2)几何BA,其细化关键帧姿势和相关联的特征映射点,以及(3)姿势图优化,以在存在闭环的情况下实现全局映射一致性。这是通过将基于特征的操作限制为来自直接测距模块的边缘化关键帧来实现的。对两个基准数据集的详尽评估表明,我们的系统在整体精度和鲁棒性方面优于最先进的单目测距和SLAM系统。
We propose a novel semi-direct approach for monocular simultaneouslocalization and mapping (SLAM) that combines the complementary strengths ofdirect and feature-based methods.The proposed pipeline loosely couples directodometry and feature-based SLAM to perform three levels of paralleloptimizations: (1) photometric bundle adjustment (BA) that jointly optimizesthe local structure and motion, (2) geometric BA that refines keyframe posesand associated feature map points, and(3) pose graph optimization to achieveglobal map consistency in the presence of loop closures.This is achieved inreal-time by limiting the feature-based operations to marginalized keyframesfrom the direct odometry module.Exhaustive evaluation on two benchmarkdatasets demonstrates that our system outperforms the state-of-the-artmonocular odometry and SLAM systems in terms of overall accuracy androbustness.论文代码开源:松耦合半直接单目SLAM(Loosely-Coupled Semi-Direct Monocular SLAM) R32Tc6t8ZJGm2f2I.jpg
URL地址:https://arxiv.org/abs/1807.10073v1     ----pdf下载地址:https://arxiv.org/pdf/1807.10073v1    ----         ----github下载地址:https://github.com/sunghoon031/LCSD_SLAM    ----    论文代码开源:松耦合半直接单目SLAM(Loosely-Coupled Semi-Direct Monocular SLAM)请注意该人工智能论文代码开源在github,大部分是python写的,框架可能是tensorflow或者pytorch,keras,至于具体是哪一个没有完全测试。
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

admin当前离线
管理员

查看:2207 | 回复:0

快速回复 返回顶部 返回列表