Grid-based multi-object tracking with Siamese CNN based appearance edge and access region mechanism
- PDF / 6,467,241 Bytes
- 19 Pages / 439.642 x 666.49 pts Page_size
- 107 Downloads / 198 Views
Grid-based multi-object tracking with Siamese CNN based appearance edge and access region mechanism Longtao Chen1 · Jing Lou2 · Fenglei Xu1 · Mingwu Ren1 Received: 27 November 2018 / Revised: 25 February 2019 / Accepted: 9 May 2019 / © Springer Science+Business Media, LLC, part of Springer Nature 2019
Abstract Receiving growing attention for its various applications during the last few years, multiobject tracking remains a complex and challenging problem. Conventional grid-based tracking method is an efficient and effective method to tackle multi-object tracking, whose performance can be further boosted by intuitively taking into account the appearance similarity information yet. Therefore, we introduce appearance similarity edge into the grid-based method, where a Siamese network is utilized to produce the proposed similarity edge. In addition, we build a grid model with hexagonal cells and propose an access region mechanism including accessible area definition and an automatic-generation approach for entrance/exit grids. Since our tracking framework follows ’tracking-bydetection’ paradigm, the corresponding detection information is available to be integrated into access region mechanism, which will facilitate appropriate grid modeling. We verify the proposed Siamese network based appearance edge and access region mechanism through the experiments on some popular datasets like PETS-09, KITTI. Keywords Grid-based tracking · Globally optimization tracking · Multi-object tracking · Siamese neural networks
1 Introduction Tracking multiple objects is a significant problem for its wider range of applications like visual surveillance, activity analysis [24, 28, 42], autonomous driving [15] and robot navigation [13, 14]. However, there are many difficulties of multi-object tracking including occlusions, unstable appearance and mutual exclusion. Although a lot of algorithms have been developed, multi-object tracking is still challenging.
Mingwu Ren
[email protected] 1
School of Computer Science and Engineering, Nanjing University of Science and Technology, Xiaolingwei 200, Nanjing, 210094, China
2
School of Information Engineering, Changzhou Institute of Mechatronic Technology, Changzhou, Jiangsu, 213164, China
Multimedia Tools and Applications
The most popular paradigm at present is tracking-by-detection, where a detector is used to detect objects in each time step and then those objects are associated into trajectories. Tracking-by-detection paradigm helps to resist divergence and deal with the above-mentioned problems. Therefore, we employ tracking-by-detection paradigm. As an efficient method for multi-object tracking, grid-based tracking has been extensively studied in several recent works such as [3, 4, 7]. This method first discretizes physical area of interest and time into a regular grid network. Then the data association problem is converted into an integer linear programming problem while considering it with a grid-based network flow model. This discretization scheme helps to significantly reduce the
Data Loading...
