The original real-time compressive tracking use a random matrix to get the appearance model based on features extracted form the image feature space performs well in most scene. However, when the object is low-grain, low-resolution, or small, a fixed size sparse measurement matrix is not sufficient enough to preserve the structure of the image of the object. In this work, we propose a Dynamic Compressive Tracking algorithm that employs adaptive random projections that preserve the structure of the image feature space of objects during tracking. The proposed tracker automatically estimates and ranks the amount of random feature projections of the object in the compressive domain. Extensive experimental results, on challenging public available data sets shows, that the proposed dynamic compressible tracking algorithm outperforms conventional compressive tracker, and it is comparable to the state-of-the-art tracking methods.
Chen, T, Zhang, Y, Yang, T & Sahli, H 2013, Dynamic Compressive Tracking. in Proceedings of International Conference on Advances in Mobile Computing & Multimedia(MoMM '13). ACM, pp. 518, International Conference on Advances in Mobile Computing & Multimedia, Vienna, Austria, 2/12/13.
Chen, T., Zhang, Y., Yang, T., & Sahli, H. (2013). Dynamic Compressive Tracking. In Proceedings of International Conference on Advances in Mobile Computing & Multimedia(MoMM '13) (pp. 518). ACM.
@inproceedings{69d87a7cda354b3aaa7f5ca18e429b34,
title = "Dynamic Compressive Tracking",
abstract = "The original real-time compressive tracking use a random matrix to get the appearance model based on features extracted form the image feature space performs well in most scene. However, when the object is low-grain, low-resolution, or small, a fixed size sparse measurement matrix is not sufficient enough to preserve the structure of the image of the object. In this work, we propose a Dynamic Compressive Tracking algorithm that employs adaptive random projections that preserve the structure of the image feature space of objects during tracking. The proposed tracker automatically estimates and ranks the amount of random feature projections of the object in the compressive domain. Extensive experimental results, on challenging public available data sets shows, that the proposed dynamic compressible tracking algorithm outperforms conventional compressive tracker, and it is comparable to the state-of-the-art tracking methods.",
keywords = "Scene analysis,, tracking",
author = "Ting Chen and Yanning Zhang and Tao Yang and Hichem Sahli",
year = "2013",
language = "English",
pages = "518",
booktitle = "Proceedings of International Conference on Advances in Mobile Computing & Multimedia(MoMM '13)",
publisher = "ACM",
note = "International Conference on Advances in Mobile Computing & Multimedia, MoMM2013 ; Conference date: 02-12-2013 Through 04-12-2013",
}