VP-SLAM A Monocular Real-time Visual SLAM with Points Lines and Vanishing Points Andreas Georgis

2025-05-06 2 0 645.86KB 5 页 10玖币

侵权投诉

VP-SLAM: A Monocular Real-time Visual SLAM

with Points, Lines and Vanishing Points

Andreas Georgis

School of ECE, NTUA

Athens, Greece

georgisandreas@gmail.com

Panagiotis Mermigkas

School of ECE, NTUA

Athens, Greece

p.mermigkas@gmail.com

Prof. Petros Maragos

School of ECE, NTUA

Athens, Greece

maragos@cs.ntua.gr

Abstract—Traditional monocular Visual Simultaneous Local-

ization and Mapping (vSLAM) systems can be divided into three

categories: those that use features, those that rely on the image

itself, and hybrid models. In the case of feature-based methods,

new research has evolved to incorporate more information from

their environment using geometric primitives beyond points, such

as lines and planes. This is because in many environments,

which are man-made environments, characterized as ”Manhattan

world”, geometric primitives such as lines and planes occupy

most of the space in the environment. The exploitation of these

schemes can lead to the introduction of algorithms capable of

optimizing the trajectory of a Visual SLAM system and also

helping to construct an exuberant map. Thus, we present a real-

time monocular Visual SLAM system that incorporates real-time

methods for line and VP extraction, as well as two strategies

that exploit vanishing points to estimate the robot’s translation

and improve its rotation.Particularly, we build on ORB-SLAM2,

which is considered the current state-of-the-art solution in terms

of both accuracy and efﬁciency, and extend its formulation to

handle lines and VPs to create two strategies the ﬁrst optimize

the rotation and the second reﬁne the translation part from the

known rotation. First, we extract VPs using a real-time method

and use them for a global rotation optimization strategy. Second,

we present a translation estimation method that takes advantage

of last-stage rotation optimization to model a linear system.

Finally, we evaluate our system on the TUM RGB-D benchmark

and demonstrate that the proposed system achieves state-of-the-

art results and runs in real time, and its performance remains

close to the original ORB-SLAM2 system.

I. INTRODUCTION

Visual SLAM (vSLAM) systems try to estimate a robot’s

location based on the multi-view geometry of the scene,

combined with computer vision algorithms, while generating

a 3D map of the environment. It is a critical tool for 3D

reconstruction, image reﬁnement, 3D holographic application,

visual place recognition, AR/VR reality, and autonomous

vehicles like micro air vehicles (MAVs). Various vSLAM

approaches have been created based on various sensors such

as single camera, stereo camera, RGB-D camera and event-

camera. Moreover, feature-based approaches have traditionally

attracted the most attention, since, they rely on well-suited

computer vision algorithms to extract features and are more

resistant to changes in light than direct methods. However, in

low-textured or man-made environments where the extracted

point features are not well-distributed or sufﬁcient, incorporat-

ing other geometric primitives from multi-view geometry into

the system, such as lines planes, or VPs (Vanishing Points)

can boost the robustness of these systems. [1], [2], [3], [4].

On the other hand, most applications in practice work on

certain scenarios, such as man-made environments.In these

environments, to boost the performance of the vSLAM system,

the hypothesis of the Manhattan World (MW) is used [5]. The

MW, is a man-made environment with signiﬁcant structural

regularity, and with most areas of surrounding environment

being described as a box world with three mutual orthogonal

dominant directions. As a result, each MW is associated with

a frame, which is denoted as Manhattan Frame (MF) and can

be inferred from the VPs, which is the intersection of the

image projections of the 3D parallel lines in MW. As a result,

employing the VPs can lead to a reduction in pose drift [6],

[7].

Motivated by the insights made above, we present a vSLAM

system that integrates simple computer vision algorithms for

extracting lines, and VPs to reduce the drift in pose and

optimize it. More speciﬁcally, the main contributions of this

paper are summarized below::

•Employ real-time computer vision algorithms for extract-

ing lines and VPs..

•An optimization strategy of absolute rotation based on

VPs.

•A simple linear system for estimate translation.

II. RELATED WORK

Next we brieﬂy review the related works on vSLAM system

with the focus on features and on leveraging the structural

regularity to improve system performance. First, we have

ORB-SLAM2 [8] which is a popular feature-based monocular

vSLAM system, that extends the multi-threaded and keyframe-

based architecture of PTAM [9]. It uses ORB features, builds a

co-visibility graph and performs loop closing and localization

tasks.To improve the robustness of point-based methods, the

authors in [9] extracted lines from the environment and pro-

pose an algorithm to integrate them into a monocular Extended

Kalman Filter SLAM system (EKF-SLAM). Finally, in PL-

SLAM [1] points and lines extracted concurrently into a point-

based system.

On the other hand, we have vSLAM systems based on

the MW assumption, like [6] which proposes a 2 stage MF

tracking to estimate the rotation of the pose based on the

arXiv:2210.12756v2 [cs.RO] 28 Oct 2022

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

VP-SLAM:AMonocularReal-timeVisualSLAMwithPoints,LinesandVanishingPointsAndreasGeorgisSchoolofECE,NTUAAthens,Greecegeorgisandreas@gmail.comPanagiotisMermigkasSchoolofECE,NTUAAthens,Greecep.mermigkas@gmail.comProf.PetrosMaragosSchoolofECE,NTUAAthens,Greecemaragos@cs.ntua.grAbstractTraditionalmonocula...

展开>> 收起<<

VP-SLAM A Monocular Real-time Visual SLAM with Points Lines and Vanishing Points Andreas Georgis.pdf

共5页,预览1页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

VP-SLAM A Monocular Real-time Visual SLAM with Points Lines and Vanishing Points Andreas Georgis

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: