Abstract:
Sparse sampling and sparse matrices are found to be able to substantially improve object tracking in videos. This paper reports a comparison study on state-of-the-art tra...Show MoreMetadata
Abstract:
Sparse sampling and sparse matrices are found to be able to substantially improve object tracking in videos. This paper reports a comparison study on state-of-the-art tracking algorithms, including MOSSE, FCT and CSK. While many consider generic tracking problems in which targets, mostly cars and humans, are moving in various settings, we consider motorcycles in this study, or more precisely the license plates on motorcycles. Motorcycles are one the most popular transportation means in metropolitan areas, especially in Asia, However, very limited research has been done on the tracking of motorcycle license plates, which is a crucial step for license plate recognition. This study has three contributions: 1) Comparison of latest tracking algorithms for motorcycle license plates; 2) Determination of appropriate settings of the selected algorithms to highlight what needs to be considered when applying these algorithms tracking specific objects; 3) A benchmark database with videos collected under various conditions is offered for performance evaluation and technical advancement.
Date of Conference: 11-12 July 2016
Date Added to IEEE Xplore: 04 August 2016
ISBN Information: