Vehicle re-identification using multi-task deep learning network and spatio-temporal model
- PDF / 1,896,365 Bytes
- 17 Pages / 439.642 x 666.49 pts Page_size
- 83 Downloads / 149 Views
Vehicle re-identification using multi-task deep learning network and spatio-temporal model Jinjia Peng1 · Yun Hao1 · Fengqiang Xu1 · Xianping Fu1,2 Received: 4 June 2018 / Revised: 6 May 2020 / Accepted: 13 July 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Vehicle re-identification (re-ID) plays an important role in the automatic analysis of the increasing urban surveillance videos and has become a hot topic in recent years. Vehicle re-ID aims at identifying vehicles across different cameras. However, it suffers from the difficulties caused by various viewpoint of vehicles, diversified illuminations, and complicated environments. In this paper, a two-stage vehicle re-ID framework is proposed to address these challenges, which contains a feature extraction module for achieving discriminative features and a spatial-temporal re-ranking module to improve the accuracy of vehicle re-ID task. Firstly, a multi-task deep network that integrates identity predicting network, attribute recognition network and verification network is adopted to learn discriminate features. Secondly, a spatio-temporal model is built to re-rank the appearance information measurement results, which utilizes the spatio-temporal relationship to increase constraints of the images. Moreover, to facilitate progressive vehicle re-ID research, experiments are conducted on both the VeRi-776 dataset and VehicleID dataset. Both the proposed multi-task feature extraction module and spatio-temporal model achieve considerable improvements. Keywords Multi-task network · Spatial-temporal model · Vehicle re-identification
Xianping Fu
[email protected] Jinjia Peng [email protected] Yun Hao [email protected] Fengqiang Xu [email protected] 1
College of Information and Science Technology, Dalian Maritime University, Dalian, Liaoning, 116021, China
2
Pengcheng Laboratory, Shenzhen, Guangdong, 518055, China
Multimedia Tools and Applications
1 Introduction With the increasing growth of surveillance videos, accurate vehicle re-ID is of great significance in searching for a target vehicle from a large number of image datasets. At the beginning of the vehicle re-ID, most methods try to solve it by various of different sensors [3, 5, 16, 20]. However it usually needs extra cost of hardware. In addition, license plate is naturally regarded as a unique ID of a vehicle. With the continuous development of license plate recognition technology [9, 29], it has already been used widely in transportation management applications. However, with the limitation of the diversified illuminations or the limitation of the resolution of remote vehicle images, license plates often could not be captured or occluded. Moreover, different with person re-ID task that always match the features obtained by traditional methods to predict whether the persons are the same one, traditional features extracted from vehicle images are very sensitive to the fickle environment. Hence, there is few attempts to solve vehicle re-ID by traditional features. Ins
Data Loading...
