Learning Local-Global Multiple Correlation Filters for Robust Visual Tracking with Kalman Filter Redetection.

Jianming Zhang,Jin Wang,Hehua Liu,Yang Liu

doi:10.3390/s21041129

Jianming Zhang, Jin Wang + Show 2 more

Open Access

https://doi.org/10.3390/s21041129

Copy DOI

Abstract

Visual object tracking is a significant technology for camera-based sensor networks applications. Multilayer convolutional features comprehensively used in correlation filter (CF)-based tracking algorithms have achieved excellent performance. However, there are tracking failures in some challenging situations because ordinary features are not able to well represent the object appearance variations and the correlation filters are updated irrationally. In this paper, we propose a local–global multiple correlation filters (LGCF) tracking algorithm for edge computing systems capturing moving targets, such as vehicles and pedestrians. First, we construct a global correlation filter model with deep convolutional features, and choose horizontal or vertical division according to the aspect ratio to build two local filters with hand-crafted features. Then, we propose a local–global collaborative strategy to exchange information between local and global correlation filters. This strategy can avoid the wrong learning of the object appearance model. Finally, we propose a time-space peak to sidelobe ratio (TSPSR) to evaluate the stability of the current CF. When the estimated results of the current CF are not reliable, the Kalman filter redetection (KFR) model would be enabled to recapture the object. The experimental results show that our presented algorithm achieves better performances on OTB-2013 and OTB-2015 compared with the other latest 12 tracking algorithms. Moreover, our algorithm handles various challenges in object tracking well.

Highlights

With the development of artificial intelligence, sensor networks nodes equipped with more sophisticated sensing units such as cameras have become ubiquitous in cities
In order to train high-quality correlation filter (CF), it is necessary for the tracking algorithm to select an appropriate descriptor or a combination of multiple feature descriptors including a histogram of oriented gradient features (HoG) [8], color names (CN) features [9], Point of Interest features [10], Haar-like rectangular features [11], superpixel features [12], etc
This paper proposes a local–global multiple correlation filters-based object tracking algorithm

Summary

Introduction

With the development of artificial intelligence, sensor networks nodes equipped with more sophisticated sensing units such as cameras have become ubiquitous in cities. Camera-based sensor networks have been widely used in intelligent transportation and smart cities. Building a target appearance model with strong representation capability is one of the keys to achieving the high precision and robustness of object tracking. According to the different categories of target appearance representation models used, the methods of visual object tracking can be divided into two types: generative methods [1,2] and discriminative methods [3,4]. Discriminative correlation filter (DCF)-based trackers [5,6,7] have gained more and more attention from researchers due to their excellent performance. The extraction method of the target appearance features mostly determines the performance of the DCF-based tracking algorithm. This work uses a local–global multiple correlation filters model constructed by convolutional features and hand-crafted features to improve tracking performance

Methods

Results

Conclusion