Object tracking framework with Siamese network and re-detection mechanism

Daqun Li,Yi Yu,Xiaolin Chen

doi:10.1186/s13638-019-1579-x

Abstract

To improve the deficient tracking ability of fully-convolutional Siamese networks (SiamFC) in complex scenes, an object tracking framework with Siamese network and re-detection mechanism (Siam-RM) is proposed. The mechanism adopts the Siamese instance search tracker (SINT) as the re-detection network. When multiple peaks appear on the response map of SiamFC, a more accurate re-detection network can re-determine the location of the object. Meanwhile, for the sake of adapting to various changes in appearance of the object, this paper employs a generative model to construct the templates of SiamFC. Furthermore, a method of template updating with high confidence is also used to prevent the template from being contaminated. Objective evaluation on the popular online tracking benchmark (OTB) shows that the tracking accuracy and the success rate of the proposed framework can reach 79.8% and 63.8%, respectively. Compared to SiamFC, the results of several representative video sequences demonstrate that our framework has higher accuracy and robustness in scenes with fast motion, occlusion, background clutter, and illumination variations.

Highlights

Object tracking is a critical issue in the field of computer vision, which exits in a wide area of applications including video surveillance, human-computer interaction, intelligent traffic monitoring, and the military, to name a few
We compare our tracker Siamese network and re-detection mechanism (Siam-RM) with nine state-of-the-art trackers, including Fully-convolutional Siamese networks (SiamFC) [35], Siamese instance search tracker (SINT) [36], CFNet [39], DCFNet [60], LCT [61], LMCF [54], and real-time algorithms based on correlation filters (Staple [62], kernelized correlation filter (KCF) [21], DSST [22])
5 Conclusion In this paper, we aim at improving the deficient tracking ability of SiamFC in complex scenes with fast motion and similar interference

Summary

Introduction

Object tracking is a critical issue in the field of computer vision, which exits in a wide area of applications including video surveillance, human-computer interaction, intelligent traffic monitoring, and the military, to name a few. After the arbitrary object in the first frame of the video sequence is given, how to precisely locate its position in the subsequent frames is the central problem of object tracking. Object features are manually defined or combined [6,7,8,9,10,11]. Most frameworks for object tracking can be divided into generative methods [12,13,14,15] and discriminative methods [16,17,18,19,20,21,22,23,24,25]

Objectives

Methods

Results

Conclusion