Color name TLD1
Defeng Sun, Bin Dai, Chuanxiang Li, Liang Xiao · 2015
To achieve better performance of visual tracking, an improved TLD tracking algorithm was proposed. Firstly, the performance of objects detection classifiers was improved by the usage of the color attributes of objects. The accuracy of objects detector was boosted by using the color attributes of initial object, which was labeled manually. Secondly, Kalman Filter is adopted to estimate localization of candidate objects. Then, with the help of estimated candidate localizations, the research space of detectors and the complexity of computation were reduced. The improved TLD, called color name TLD, was tested on many Benchmark sequences and the experimental results show that the performance and accuracy outperform other publicly available tracking methods under the different scenes.