Accurate estimation of feature points based on individual projective plane in video sequence
- PDF / 2,706,059 Bytes
- 13 Pages / 595.276 x 790.866 pts Page_size
- 18 Downloads / 166 Views
ORIGINAL ARTICLE
Accurate estimation of feature points based on individual projective plane in video sequence Huajun Liu1 · Shiran Tang1 · Dian Lei1 · Qing Zhu2 · Haigang Sui3 · Gaojian Zhang1 · Chao Li1
© Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract The stability and quantity of feature matching in video sequence is one of the key issues for feature tracking and some relevant applications. The existing matching methods are based on feature detection, which is usually affected by illumination conditions, noise or occlusions, and this will directly influence matching results. In this paper, we propose an accurate prediction method for interest point estimation in video sequence by extracting the stable mapping for each undetected point in its suitable projective plane, which is based on coplanar feature points that have already been detected in adjacent frames. The proposed prediction method breaks the limitation of the previous approaches that largely rely on feature detection. Our experiments show that our method not only predicts features accurately, but also enriches the correspondences, which prolongs the track length of features. Keywords Feature point prediction · Homography · Projective plane · Video sequence
1 Introduction Stable feature detection and correct matching are the premise of obtaining a long feature tracking in video sequence, which
B B B
Huajun Liu [email protected] Shiran Tang [email protected] Dian Lei [email protected] Qing Zhu [email protected] Haigang Sui [email protected] Gaojian Zhang [email protected] Chao Li [email protected]
1
School of Computer Science, Wuhan University, Wuhan, China
2
Faculty of Geosciences and Environmental Engineering, Southwest Jiaotong University, Chengdu, China
3
State Key Laboratory of Information Engineering in Surveying Mapping and Remote Sensing, Wuhan University, Wuhan, China
directly affects the precision of relevant applications, such as bundle adjustment [30], image registration [37], panorama production [33], and 3D reconstruction [11,15]. Feature detection is the precondition of the existing matching methods. However, the projective points of the same 3D point may not be detected stably in adjacent frames due to illumination change and other factors, which will result in fewer correspondences. Therefore, it is one of the key issues to find more stable feature points for consecutive matching in video frames. The feature matching approaches that are widely used are based on feature detection, such as Kanade–Lucas–Tomasi tracking [17] based on Harris [12] corner detector, matching with descriptors based on scale-invariant feature transform (SIFT) [23], oriented FAST or rotated BRIEF (ORB) [28]. However, when there exists illumination change, noise or repeated texture, traditional matching methods which only use a single type of descriptors by a strict threshold cannot acquire enough correct correspondences. In order to enrich correct matches, Zhang et al. presented an improved two-pass matching strategy w
Data Loading...
