Correlation filter based tracking methods are the core component of most trackers which achieve the excellent performance in term of the accuracy and robustness in visual tracking. However, there are still lots of challenging situations, such as occlusion or illumination, which confines and limits the performance of trackers. To cope with the above problems, in this paper, we suggest an effective tracking method via part-based strategy. Compared with the conventional tracking algorithms based on correlation filter, our tracker employs the novel strategy to validate and estimate the target’s final position, avoiding merely utilizing the maximum response in the response map as the target position which is often prone to drift away from the target. In addition, to effectively deal with occlusion, we divide the sample into multiple parts. When the sample is partly occluded, the visible part can still provide effective clues for tracking, ensuring the robustness of tracker. A large number of conveys are conducted on the public databases, and experimental results show that the proposed algorithm has obvious performance improvement in the case of dealing with target occlusion, and the real-time performance is also pretty good.
Visual object tracking has become increasingly popular in the community due to its application and research significance. However, occlusion is one of the major factors that seriously impact the tracking performance in visual tracking. To address this issue, in this paper, we propose a novel nonlocal correlation filter based tracking method. Our proposed tracker effectively exploits the explicit coupled mechanism which depends on the global filter and several local part filters, and efficiently employs the spatial geometric constraints among the global object and local patches of object for preserving the structure of object. Compared with other existing correlation filter based trackers, our proposed tracking method has three advantages: (1) To ensure complete representation to the target candidate, we learn the correlation filers from not only the global sample but also local sample parts. The global based filter guarantees the overall accuracy of the tracked object, while the local based filters reserve the details of tracking object to cope with the challenging cases like occlusion or deformation. In addition, an effective and adaptive selection mechanism is proposed to select the most distinctive and discriminative parts for tracking, which avoids unnecessary computing burden caused by tracking all parts and simultaneously improves the robustness of the tracker. (2) Through adaptively weighting the global sample and each local part of samples, the integration mechanism puts more emphasis on visible parts and eliminates the impacts by occluded parts for further improving the tracking robustness. (3) Different from other trackers by searching for the predefined scale pyramid, we propose a simple yet effective scale estimation strategy which can accurately calculate the current scale of the tracking target. For verifying our method, we conduct extensive qualitative and quantitative experiments on challenging benchmark image sequences. Experiment results demonstrate that our proposed method performs favorably against several state-of-the-art trackers.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.