Real-Time Visual Tracking: Promoting the Robustness of Correlation Filter Learning
Correlation filtering based tracking model has received lots of attention and achieved great success in real-time tracking, however, the lost function in current correlation filtering paradigm could not reliably response to the appearance changes caused b
- PDF / 7,426,099 Bytes
- 17 Pages / 439.37 x 666.142 pts Page_size
- 79 Downloads / 212 Views
Department of EECS, University of Kansas, Lawrence, KS 66045, USA [email protected], [email protected] 2 Department of ECE, Boston University, Boston, MA 02215, USA [email protected] 3 China Unicom Research Institute, Beijing 100032, China [email protected] 4 Department of EE, Tsinghua University, Beijing 100084, China [email protected]
Abstract. Correlation filtering based tracking model has received lots of attention and achieved great success in real-time tracking, however, the lost function in current correlation filtering paradigm could not reliably response to the appearance changes caused by occlusion and illumination variations. This study intends to promote the robustness of the correlation filter learning. By exploiting the anisotropy of the filter response, three sparsity related loss functions are proposed to alleviate the overfitting issue of previous methods and improve the overall tracking performance. As a result, three real-time trackers are implemented. Extensive experiments in various challenging situations demonstrate that the robustness of the learned correlation filter has been greatly improved via the designed loss functions. In addition, the study reveals, from an experimental perspective, how different loss functions essentially influence the tracking performance. An important conclusion is that the sensitivity of the peak values of the filter in successive frames is consistent with the tracking performance. This is a useful reference criterion in designing a robust correlation filter for visual tracking. Keywords: Visual tracking · Correlation filtering · Sparsity regularization · Loss function · Robustness
1
Introduction
In recent years, there is a significant interest in correlation filtering based tracking. Under this paradigm, a correlation filter is efficiently learned online from previously obtained target regions, and the target is located according to the magnitude of the filter response over a large number of target candidates. The main strength of this paradigm is its high computational efficiency, because the target and the candidate regions can be represented in frequency domain and manipulated by fast Fourier transform (FFT), which yields O (n log n) c Springer International Publishing AG 2016 B. Leibe et al. (Eds.): ECCV 2016, Part VIII, LNCS 9912, pp. 662–678, 2016. DOI: 10.1007/978-3-319-46484-8 40
Real-Time Visual Tracking
663
√ √ computational complexity for a region of n × n pixels. For this reason, extensive real-time trackers [1–9] have been proposed within the correlation filtering paradigm. Specifically, a correlation filter is learned from previously obtained target regions to approximate an expected filter response, such that the peak of the response is located at the center of the target region. The response used in previous methods is often assigned to be of Gaussian shaped, which is treated as a continuous version of an impulse signal. For this reason, the learned filter is encouraged to produce Gaussian shaped response. The candidate region with the strongest filter
Data Loading...
