基于概率解耦自适应运动估计和三维SSC的鲁棒立体视觉测距1
需积分: 0 53 浏览量
2022-08-03
21:51:55
上传
评论
收藏 3.63MB PDF 举报
2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2018.2886824, IEEE Access
Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.DOI
Robust Stereo Visual Odometry Based
on Probabilistic Decoupling Ego-motion
Estimation and 3D SSC
YAN WANG
1
, HUI-QI MIAO
1
, and LEI GUO
1
1
School of Automation Science and Electrical Engineering, Beihang University, Beijing, China 100191
Corresponding author: Yan Wang (e-mail: 18811409162miao@gmail.com).
This work was supported by the National Nature Science Foundation of China [Grant numbers 61573019, 61751302, 61320106010,
61627810].
ABSTRACT The outliers caused by noise and mismatching severely restrict the precision of visual
odometry. Moreover, the dynamic environment is also a crucial element that decreases the robustness
of the systems. This paper presents a robust stereo visual odometry by decoupled ego-motion estimation
based on probabilistic matches and rejecting the outliers of dynamic objects through motion segmentation.
Fast ZNCC method, based on local sum table and partition upper bound schemes, is presented for
selecting probabilistic matches while keeping run-time efficiency. The selection of multi-correspondences
can avoid mismatching of corresponding points. In consideration of noise interference, the essential matrix is
computed in probabilistic framework to estimate the initial value of rotation matrix without estimated depth
errors involved. Then in order to estimate pose robustly in dynamic environment, a modified SSC method is
discussed which aims to cluster the tracked 3D points cloud to avoid errors caused by affine transformation.
And the non-negative constraint makes the method suitable for fast moving camera. The proposed 3D-
SSC method removes the outliers belonging to dynamic objects effectively . Finally, the detected inliers
and depths are employed to estimate translation matrix and refine rotation matrix. The proposed method is
evaluated on KITTI benchmark and compared with state of art methods. The results show that our method is
more robust as it can detect outliers more accurately in dynamic environments and achieve higher precision
in motion estimation.
INDEX TERMS Stereo visual odometry, probabilistic matches, decoupling estimation, 3D SSC.
I. INTRODUCTION
I
N recent years, visual odometry (VO) and visual SLAM
have played an immensely important role on autonomous
driving [1], [2]. With the abundant information provided via
vision, the autonomous system can generate self-localization
measurements. Most odometry methods preform registration
between the current image and a previous reference, in which
the estimated transformations between these images are as-
sumed to originate from the camera motion [3]. However,
almost all the techniques are built under the assumption of
static environments, which usually cannot be satisfied in the
real world. Dynamic objects which violate this assumption
will seriously influence the precision of estimation.
Improving the performance of visual odometry in dy-
namic environments is an important and desirable problem,
especially for vehicles. For cameras equipped on vehicles
capturing dynamic scenes, both static and dynamic scene
parts appear to be moving [4]. It is seldom the case that
vehicles operate in strictly static environments. Therefore,
the moving objects in environments will significantly impact
the accuracy of estimation.
In addition, the ego-motion of mobile vehicles consists
of the rotation R and the translation t. They are estimated
and integrated together in most VO approaches, which are
prone to drift. From an application perspective, the location
information of the vehicle is crucial information supported by
the odometry. In the KITTI [5] vision benchmark scoreboard,
the translation error is regarded as the exclusive factor for
ranking, and the rotation error is displayed for reference
only. Nevertheless, rotation errors have greater influence than
translation errors on final location during the cumulation
of errors in dead reckoning process such as odometry. The
translation is reliant on the depth in contrast to the rotation.
VOLUME 4, 2016 1
评论0