Int J Performability Eng ›› 2018, Vol. 14 ›› Issue (2): 386-395.doi: 10.23940/ijpe.18.02.p20.386395

• Original articles • Previous Articles    

Target Tracking based on KCF Combining with Spatio-Temporal Context Learning

Aili Wanga, Zhennan Yanga, Yushi Chenb, and Yuji Iwahoric   

  1. aHigher Education Key Lab for Measure& Control Technology and Instrumentations of Heilongjiang, Harbin University of Science and Technology, Harbin, 150080, China
    bSchool of Electronics and Information Engineering, Harbin Institute of Technology, Harbin, 150001, China
    cDept of Computer Science, Chubu University, Aichi, Japan

Abstract:

Most target tracking is based on a lot of samples training to build the model of the target, which is then carried on the tracking processing. This will need to choose a lot of tracked target samples for learning and training. However, there are all kinds of deformation of the training samples, including variety of light and scale, and so on, causing the long computation time, high computational complexity, and less robustness. The traditional kernel correlation filtering (KCF) tracking is through online learning of the first frame in the target vide. It then uses cyclic matrix to strengthen samples robustness, reducing the complexity of the calculation and time. But, the traditional KCF nuclear is unsatisfactory used for complex scenarios and quick treatment. In this paper, under the framework of the KCF, the target context information is introduced to make the tracking have better robustness and a better effect to deal with complex scenarios.