A Survey of Vehicle Re-Identification Based on Deep Learning

Hongbo Wang,Na Chen,Jiaying Hou

doi:10.1109/access.2019.2956172

Abstract

Vehicle re-identification is one of the core technologies of intelligent transportation systems, and it is crucial for the construction of smart cities. With the rapid development of deep learning, vehicle re-identification technologies have made significant progress in recent years. Therefore, making a comprehensive survey about the vehicle re-identification methods based on deep learning is quite indispensable. There are mainly five types of deep learning-based methods designed for vehicle re-identification, i.e. methods based on local features, methods based on representation learning, methods based on metric learning, methods based on unsupervised learning, and methods based on attention mechanism. The major contributions of our survey come from three aspects. First, we give a comprehensive review of the current five types of deep learning-based methods for vehicle re-identification, and we further compare them from characteristics, advantages, and disadvantages. Second, we sort out vehicle public datasets and compare them from multiple dimensions. Third, we further discuss the challenges and possible research directions of vehicle re-identification in the future based on our survey.

Highlights

In recent years, the development of technology in the field of computer vision and the breakthrough of technology in the field of Internet of Things promote the realization of smart city concept [1]
Some methods based on traditional machine learning are introduced, focusing on methods based on deep learning which include methods based on local features, methods based on representation learning, methods based on metric learning, methods based on unsupervised learning, methods based on attention mechanism and other vehicle re-identification methods
The advantages of methods based on local features are reflected in it can capture unique visual clues conveyed by local areas and improve the perception of nuance, which helps a lot to distinguish between different vehicles and improve the accuracy of vehicle re-identification

Summary

INTRODUCTION

The development of technology in the field of computer vision and the breakthrough of technology in the field of Internet of Things promote the realization of smart city concept [1]. It has a vital role in applications such as live monitoring or multi-view vehicle tracking for urban surveillance, vehicle reidentification technology is crucial to the future development of the Internet of things, as well as the construction of intelligent transportation system and smart city Both pedestrians and vehicles are common objects in smart city applications, most attention has been paid to person re-identification in recent years due to the abundance of well-annotated pedestrian data, along with the historical focus of computer vision research on human faces and bodies [120]. The paper is organized as follows: Section II presents vehicle re-identification methods based on traditional machine learning and deep learning which are further categorized into five categories and gives the comparison between different methods. Multi-dimensional comparison of vehicle re-identification algorithms based on traditional machine learning

VEHICLE RE-IDENTIFICATION METHODS BASED ON TRADITIONAL MACHINE LEARNING

SUMMARY

Findings

CONCLUSION