Towards learning line descriptors from patches: a new paradigm and large-scale dataset
- PDF / 3,769,898 Bytes
- 14 Pages / 595.276 x 790.866 pts Page_size
- 17 Downloads / 215 Views
ORIGINAL ARTICLE
Towards learning line descriptors from patches: a new paradigm and large‑scale dataset Hongmin Liu1,2 · Yujie Liu2 · Miaomiao Fu2 · Yuhui Wei2 · Zhanqiang Huo2 · Yingxu Qiao2 Received: 9 May 2020 / Accepted: 19 September 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract Line feature description is important for image matching. However, its development is much slow compared to point description, and is still in the stage of manual design, which suffers from the problem of weak distinguish ability and poor robustness under complex conditions. To improve on this situation, this paper proposes to learn the line feature description based on convolutional neural network. First, a large-scale dataset consisting of about 229,000 labeled pairs of matched lines is built for training and testing. Then, a paradigm for learning the line descriptors based on the constructed line dataset is proposed. Specifically, the line is represented uniquely by the stacked mean and standard deviation patches of the support regions of those points lying on the line, which is subsequently fed into the L2Net to output the required line descriptors directly. Based on the line matching principals, the network is also trained with the triplet loss that is widely used for learning point descriptors. Experimental results for line matching and curve matching both demonstrate the superiority and effectiveness of the proposed learning-based descriptor, especially, averaged increases of 4.66 ~ 5.7% mAPs, 10.59 ~ 12.10% mAPs, 0.96 ~ 3.75% mAPs and 3.73% mAP on testing subset, Oxford dataset, line dataset and curve dataset are obtained compared to handcrafted descriptors. As an application, we apply the learned line descriptor to image stitching and also obtain good results. Keywords Line feature description · Point description · Convolutional neural network · Large-scale dataset
1 Introduction Image feature description is the basis of a wide range of computer vision applications, such as object and scene recognition [1–3], object detection [4–6], image stitching [7, 8], face recognition [9, 10], camera tracking [11], 3D reconstruction [12–14], image retrieval [15, 16], and so on. For a variety of image changes, e.g. illumination, viewpoint, rotation, scale and blur, the robustness and distinctiveness of the feature description influence the results of the applications essentially. With the success of Alexnet [17], the methods based on deep neural network have been widely adopted in various * Zhanqiang Huo [email protected] * Yingxu Qiao [email protected] 1
School of Automation and Electrical Engineering, University of Science and Technology Beijing, Beijing 100083, China
School of Computer Science and Technique, Henan Polytechnic University, Jiaozuo 454003, China
2
fields of computer vision, including feature description, more precisely, feature point description. Han et al. proposed the MatchNet [18], which consists of a deep convolutional network that extracts features from patches and a networ
Data Loading...
