Guided Matching Based on Statistical Optical Flow for Fast and Robust Correspondence Analysis
In this paper, we present a novel algorithm for reliable and fast feature matching. Inspired by recent efforts in optimizing the matching process using geometric and statistical properties, we developed an approach which constrains the search space by uti
- PDF / 1,021,254 Bytes
- 17 Pages / 439.37 x 666.142 pts Page_size
- 35 Downloads / 205 Views
AIT Austrian Institute of Technology, Vienna, Austria {josef.maier.fl,martin.humenberger, markus.murschitz.fl,oliver.zendel}@ait.ac.at 2 Vienna University of Technology, Vienna, Austria [email protected]
Abstract. In this paper, we present a novel algorithm for reliable and fast feature matching. Inspired by recent efforts in optimizing the matching process using geometric and statistical properties, we developed an approach which constrains the search space by utilizing spatial statistics from a small subset of matched and filtered correspondences. We call this method Guided Matching based on Statistical Optical Flow (GMbSOF). To ensure broad applicability, our approach works on high dimensional descriptors like SIFT but also on binary descriptors like FREAK. To evaluate our algorithm, we developed a novel method for determining ground truth matches, including true negatives, using spatial ground truth information of well known datasets. Therefore, we evaluate not only with precision and recall but also with accuracy and fall-out. We compare our approach in detail to several relevant state-ofthe-art algorithms using these metrics. Our experiments show that our method outperforms all other tested solutions in terms of processing time while retaining a comparable level of matching quality. Keywords: Image matching · Correspondence analysis · Statistical optical flow · Guided matching · Ground truth for feature matching
1
Introduction
Many modern real-time computer vision applications, such as visual odometry for autonomous driving or navigation of unmanned aerial vehicles, require not This work was funded by the Austrian Research Promotion Agency (FFG) project RoSSATA (contract #849035). Electronic supplementary material The online version of this chapter (doi:10. 1007/978-3-319-46478-7 7) contains supplementary material, which is available to authorized users. c Springer International Publishing AG 2016 B. Leibe et al. (Eds.): ECCV 2016, Part VII, LNCS 9911, pp. 101–117, 2016. DOI: 10.1007/978-3-319-46478-7 7
102
J. Maier et al.
only accurate but also fast detection and tracking of distinctive parts across several images. A well known approach to tackle this challenge is called feature matching. A feature is represented by a keypoint and its descriptor, thus, feature matching consists of keypoint detection (e.g. FAST [1]), descriptor extraction (e.g. SIFT [2] or FREAK [3]), and correspondence analysis. While similarity of descriptors is the main measure for correspondence analysis, higher speed as well as more robustness can be achieved by employing additional information such as statistical distributions of keypoints, estimated geometry, or even a priori knowledge about the scene. Impressive results have been achieved in the past two decades and a summary is given in Sect. 2. However, matching quality and especially processing speed can still be improved to broaden applicability. Thus, the first contribution of this work is a novel algorithm for fast and robust correspondence analysis (Sect. 3). We ca
Data Loading...
