# Probabilistic Background Model by Density Forests for Tracking

Daimu Oiwa (Department of Computer Science, Chubu University, Kasugai, Japan), Shinji Fukui (Department of Computer Science, Aichi University of Education, Kariya, Japan), Yuji Iwahori (Department of Computer Science, Chubu University, Kasugai, Japan), Tsuyoshi Nakamura (Department of Computer Science and Engineering, Nagoya Institute of Technology, Nagoya, Japan), Boonserm Kijsirikul (Department of Computer Engineering, Chulalongkorn University, Bangkok, Thailand) and M. K. Bhuyan (Department of Electronics and Electrical Engineering, Indian Institute of Technology Guwahati, Guwahati, India)
DOI: 10.4018/IJSI.2017040101

## Abstract

This paper proposes an approach for a robust tracking method to the objects intersection with appearances similar to a target object. The target is image sequences taken by a moving camera in this paper. Tracking methods using color information tend to track mistakenly a background region or an object with color similar to the target object since the proposed method is based on the particle filter. The method constructs the probabilistic background model by the histogram of the optical flow and defines the likelihood function so that the likelihood in the region of the target object may become large. This leads to increasing the accuracy of tracking. The probabilistic background model is made by the density forests. It can infer a probabilistic density fast. The proposed method can process faster than the authors' previous approach by introducing the density forests. Results are demonstrated by experiments using the real videos of outdoor scenes.
Article Preview

## Tracking Based On Particle Filter

The proposed method improves the method (Watanabe et al., 2012) so as to track the target object at high speed. The method (Watanabe et al., 2012) is based on the particle filter. The tracking method is described in this section.

### Outline of Proposed Method

The outline of the proposed method is as follows:

• Step 0: Initialization

• Step 1: Judging the situation under which the target object exists

• Step 2: Prediction

• Step 3: Calculating a weight of each particle after calculating likelihood

• Step 4: Estimating the state variables by the weighted mean of the state variables of all particles

• Step 5: Filtering by weighted resampling with replacement

The steps from STEP 1 to STEP 5 are processed at each frame.

The processes from STEP 0 to STEP 3 are described in the following subsections.

## Complete Article List

Search this Journal:
Reset
Open Access Articles: Forthcoming
Volume 5: 4 Issues (2017): 2 Released, 2 Forthcoming
Volume 4: 4 Issues (2016)
Volume 3: 4 Issues (2015)
Volume 2: 4 Issues (2014)
Volume 1: 4 Issues (2013)
View Complete Journal Contents Listing