Volume XL-4/W5
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XL-4/W5, 49-53, 2015
https://doi.org/10.5194/isprsarchives-XL-4-W5-49-2015
© Author(s) 2015. This work is distributed under
the Creative Commons Attribution 3.0 License.
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XL-4/W5, 49-53, 2015
https://doi.org/10.5194/isprsarchives-XL-4-W5-49-2015
© Author(s) 2015. This work is distributed under
the Creative Commons Attribution 3.0 License.

  11 May 2015

11 May 2015

ADAPTIVE PARAMETER ESTIMATION OF PERSON RECOGNITION MODEL IN A STOCHASTIC HUMAN TRACKING PROCESS

W. Nakanishi, T. Fuse, and T. Ishikawa W. Nakanishi et al.
  • Dept. of Civil Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 1138656 Japan

Keywords: State space modelling, Human tracking, Sequential image, Adaptive parameter estimation, Close Range

Abstract. This paper aims at an estimation of parameters of person recognition models using a sequential Bayesian filtering method. In many human tracking method, any parameters of models used for recognize the same person in successive frames are usually set in advance of human tracking process. In real situation these parameters may change according to situation of observation and difficulty level of human position prediction. Thus in this paper we formulate an adaptive parameter estimation using general state space model. Firstly we explain the way to formulate human tracking in general state space model with their components. Then referring to previous researches, we use Bhattacharyya coefficient to formulate observation model of general state space model, which is corresponding to person recognition model. The observation model in this paper is a function of Bhattacharyya coefficient with one unknown parameter. At last we sequentially estimate this parameter in real dataset with some settings. Results showed that sequential parameter estimation was succeeded and were consistent with observation situations such as occlusions.