Because uniform division spatial histogram can not finely divide the data in relatively concentrated areas, it can not accurately track human faces. A new face tracking method which combines an improved spatial histogram with particle filter is proposed. In this method, non-uniform division is proposed. Histogram data in relatively concentrated areas can be divided finely, and histogram data in relatively sparse areas can be divided roughly. Simultaneously, a new re-sampling method is proposed in order to solve the "particle degradation" and "particle depletion". If many duplicate particles occur, keep a particle, remove other particles. In order to ensure that the total number of particles is N, particles must be selected randomly in the vicinity of the particles which have a large weight. Experiments show that its tracking performance is very good when target color is similar to the scene color and obstructed partly or completely, or under the complex non-linear, non-Gaussian situations.
CITATION STYLE
Yang, D., Zhang, Y., Ji, R., Li, Y., Huangfu, L., & Yang, Y. (2015). An improved spatial histogram and particle filter face tracking. In Advances in Intelligent Systems and Computing (Vol. 329, pp. 257–267). Springer Verlag. https://doi.org/10.1007/978-3-319-12286-1_26
Mendeley helps you to discover research relevant for your work.