SPECTRAL FILTER TRACKING

Abstract

Visual object tracking is a challenging computer vision task with numerous real-world applications. In this paper, we propose a simple but efficient Spectral Filter Tracking (SFT) method from the view of graph, where each candidate image region is modeled as a pixelwise grid graph. Instead of the conventional graph matching, we formulate the tracking as a plain least square regression problem of learning spectral filters on graphs to predict an optimal vertex, which indicates the center of the target. To bypass computationally expensive eigenvalue decomposition on graph Laplacian L, we parameterize spectral graph filters as a polynomial of L to aggregate local graph features according to spectral graph theory, in which Lk exactly encodes a k-hop local neighborhood of each vertex. Thus, different from the holistic regression in those correlation filter based methods, SFT can operate on localized regions around a pixel (i.e., a vertex), which can effectively reduce the influence of local variations and cluttered backgrounds. Furthermore, we observe that the correlation filter tracking may be viewed as a specific case of our proposed spectral filtering method. The implementation of SFT can simply boil down to only a few line codes, but surprisingly it beats the correlation filter based model with the same feature input, and achieves the state-of-the-art performance on OTB-2015 and VOT2016 under the same feature extraction strategy.

Existing System

Minimum Output Sum of Squared Error (MOSSE)

Proposed System

In this paper we propose a simple but efficient Spectral Filter Tracking (SFT) method by performing local filtering on candidate region. Specifically, we model the tracking problem into a graph framework by viewing each candidate region as a pixelwise grid graph. The advantage is two-fold: i) avoiding any operations of part partition or superpixel segmentation as used in those part based methods; ii) preserving some excellent properties of graph itself such as deformation/rotation invariance. The tracking target can be defined as a subgraph of the whole candidate region, and thus the tracking falls into the category of graph matching. Rather than using the conventional graph matching strategy, we formulate the tracking as a plain least square regression problem of learning spectral graph filters. Spectral filters are performed on the graph to extract robust features. To avoid expensive computation of eigenvalue decomposition on graph Laplacian matrix, spectral filters are parameterized as a polynomial of graph Laplacian matrix, in which the k-th term of Laplacian matrix exactly defines a k-localized spatial subgraph region. Consequently, spectral filtering on graph is approximated to an operation on graph Laplacian matrix. Meantime, the polynomial terms of spectral filters actually describe different-size receptive fields. By taking the terms of the Laplacian polynomial as the filter bases, we can formulate the learning of the corresponding polynomial parameters as well as feature transformation into a plain regression model. Through some theoretical analysis, we find that the correlation filter tracking may be viewed as an extreme specific case of our proposed spectral filtering method.  

Architecture

BLOCK DIAGRAM