Visual object tracking: in the simultaneous presence of scale variation and occlusion
Authors:
Open Access
- 1 January 2018
- journal article
- research article
- Published by Taylor & Francis in Systems Science & Control Engineering
- Vol. 6 (1) , 456-466
Abstract
Visual object tracking is a challenging task when the object appearance changes caused by the scale variation and the occlusion. In this paper, an object tracking algorithm is proposed which is capable of dealing with the case that the scale variation and the occlusion occur simultaneously. A kernelized correlation filter (KCF) is first learned to obtain the correlation response, whose maximum value denotes the optimal object location. In order to represent the sample better, the convolutional features are extracted from a pre-trained convolutional neural networks (CNNs). Then, the strategy of scale adaption is used to estimate the object scale during the tracking process. Subsequently, a novel re-detection model is proposed by using a support vector machine (SVM) classifier to re-find the object when the occlusion occurs. The comparison experiments are implemented on the object tracking benchmark (OTB) and the results demonstrate that the proposed tracking algorithm outperforms other state-of-the-art ones in terms of precision and success rate.Keywords
Funding Information
- National Natural Science Foundation of China (61473076)
- National Natural Science Foundation of China (61673103)
- National Natural Science Foundation of China (61873059)
- Natural Science Foundation of Shanghai Municipality
- Program for Professor of Special Appointment (Eastern Scholar) at Shanghai Institutions of Higher Learning of China
This publication has 36 references indexed in Scilit:
- Long-term correlation tracking2015
- Fully convolutional networks for semantic segmentation2015
- High-Speed Tracking with Kernelized Correlation FiltersIEEE Transactions on Pattern Analysis and Machine Intelligence, 2014
- Object Tracking With Joint Optimization of Representation and ClassificationIEEE Transactions on Circuits and Systems for Video Technology, 2014
- Adaptive Color Attributes for Real-Time Visual Tracking2014
- Robust Object Tracking via Sparse Collaborative Appearance ModelIEEE Transactions on Image Processing, 2014
- Real-Time Compressive Tracking2012
- Exploiting the Circulant Structure of Tracking-by-Detection with Kernels2012
- Discriminative Tracking by Metric Learning2010
- Support-vector networksMachine Learning, 1995