SIFT SURF算法的比较.pdf

发布时间：2022-05-29 发布人：admin 分类：说明书资料大小：7.08M 资料格式：pdf 举报版权申诉

sdudatou-2619308-2079c82b9b11f8f7cf2d55198a992611.pdf-第1页.png

第1页 / 共8页

sdudatou-2619308-2079c82b9b11f8f7cf2d55198a992611.pdf-第2页.png

第2页 / 共8页

sdudatou-2619308-2079c82b9b11f8f7cf2d55198a992611.pdf-第3页.png

第3页 / 共8页

sdudatou-2619308-2079c82b9b11f8f7cf2d55198a992611.pdf-第4页.png

第4页 / 共8页

sdudatou-2619308-2079c82b9b11f8f7cf2d55198a992611.pdf-第5页.png

第5页 / 共8页

sdudatou-2619308-2079c82b9b11f8f7cf2d55198a992611.pdf-第6页.png

第6页 / 共8页

sdudatou-2619308-2079c82b9b11f8f7cf2d55198a992611.pdf-第7页.png

第7页 / 共8页

sdudatou-2619308-2079c82b9b11f8f7cf2d55198a992611.pdf-第8页.png

第8页 / 共8页

SIFT, SURF & seasons: Appearance-based long-term localization in outdoor environments

Introduction

Feature detectors and descriptors

SURF variants

Local feature matching

The data sets

Evaluation of local feature algorithms

Experiment 1: Single-image localization

Experiment 2: Feature matching over seasons

Time consumption

Reliable localization using epipolar constraints

Epipolar constraint

Final results

Conclusions

Acknowledgments

References

Robotics and Autonomous Systems 58 (2010) 149–156 Contents lists available at ScienceDirect Robotics and Autonomous Systems journal homepage: www.elsevier.com/locate/robot SIFT, SURF & seasons: Appearance-based long-term localization in outdoor environments Christoffer Valgren∗, Achim J. Lilienthal AASS Research Centre, Department of Computer Science, Örebro University, SE-70182 Örebro, Sweden a r t i c l e i n f o a b s t r a c t Article history: Available online 22 September 2009 Keywords: Localization Scene recognition Outdoor environments 1. Introduction In this paper, we address the problem of outdoor, appearance-based topological localization, particularly over long periods of time where seasonal changes alter the appearance of the environment. We investigate a straightforward method that relies on local image features to compare single-image pairs. We first look into which of the dominating image feature algorithms, SIFT or the more recent SURF, that is most suitable for this task. We then fine-tune our localization algorithm in terms of accuracy, and also introduce the epipolar constraint to further improve the result. The final localization algorithm is applied on multiple data sets, each consisting of a large number of panoramic images, which have been acquired over a period of nine months with large seasonal changes. The final localization rate in the single-image matching, cross- seasonal case is between 80% to 95%. © 2009 Elsevier B.V. All rights reserved. Local feature matching has become an increasingly used method for comparing images. Various methods have been proposed. The Scale-Invariant Feature Transform (SIFT) by Lowe [1] has, with its high accuracy and relatively low computation time, become the de facto standard. Some attempts of further improvements to the algorithm have been made (for example PCA- SIFT by Ke and Sukthankar [2]). Perhaps the most recent, promising approach is the Speeded Up Robust Features (SURF) by Bay et al. [3], which has been shown to yield comparable or better results to SIFT while having a fraction of the computational cost [3,4]. For mobile robots, reliable image matching can form the basis for localization and loop closing detection. Local feature algorithms have been shown to be a good choice for image matching tasks on a mobile platform, as occlusions and missing objects can be handled. There are many works in relation to appearance-based global localization using local features, for example by Se et al. [5]. In particular, SIFT applied to panoramic images has been shown to give good results in indoor environments [6,7] and also to some extent in outdoor environments [8]. However, there are relatively few approaches to topological localization by matching outdoor images from different seasons. Under outdoor conditions, the appearance of the environment is inevitably altered over ∗ Corresponding author. E-mail addresses: christoffer.wahlgren@gmail.com (C. Valgren), achim@lilienthals.de (A.J. Lilienthal). URLs: http://www.aass.oru.se/Research/Learning/crwn.html (C. Valgren), http://www.aass.oru.se/Research/Learning/amll.html (A.J. Lilienthal). 0921-8890/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.robot.2009.09.010 different time scales: changing lighting conditions, shadows and seasonal changes, etc. All of these aspects make image matching very difficult.1 Some attempts have been made to match outdoor images from different seasons. Zhang and Kosecka [9] focus on recognizing buildings in images, using a hierarchical matching scheme where a ‘‘localized color histogram’’ is used to limit the search in an image database, with a final localization step based on SIFT feature matching. He et al. [10] also use SIFT features, but employ learning over time to find ‘‘feature prototypes’’ that can be used for localization. In this paper, we investigate topological localization based on local features extracted from panoramic images. Several other works rely on similar techniques to do topological mapping and localization, for example Booij et al. [7], Sagues et al. [11] and Valgren et al. [8,12]. The most recent work related to this paper is a comparative study for the localization task in indoor environments, published by Murillo et al. [13], where it is found that SURF outperforms SIFT because of its high accuracy and lower computation time. The aim of this investigation is to determine the topological localization performance that can be achieved using local features only under the conditions of large appearance changes over seasons. Our approach is to first evaluate the localization performance with SIFT and SURF on outdoor data. We then choose the local feature algorithm with the highest performance, introduce the epipolar constraint, and adjust the parameters to reach as high localization rate as possible. 1 In some cases even impossible, since a snow-covered field might not provide any discriminative features.

分享到：

赞收藏

资料库

SIFT SURF算法的比较.pdf

相关推荐

开发技术

热门标签

最新资料