2006
DOI: 10.1109/tmm.2006.884624
|View full text |Cite
|
Sign up to set email alerts
|

A Nonparametric Adaptive Tracking Algorithm Based on Multiple Feature Distributions

Abstract: This paper presents an object tracking framework based on the mean-shift algorithm, which is a nonparametric technique that uses statistical color distribution of objects. Tracking objects through highly similar-colored background is one of the problems that need to be addressed. In various cases where object and background color distributions are very similar, the color distribution obtained from single frame alone is not sufficient to track objects reliably. To deal with this problem, the proposed algorithm … Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1

Citation Types

0
5
0

Year Published

2008
2008
2018
2018

Publication Types

Select...
5
2
1

Relationship

0
8

Authors

Journals

citations
Cited by 17 publications
(5 citation statements)
references
References 18 publications
0
5
0
Order By: Relevance
“…Most multiple human tracking approaches fall into one of three categories: achieving a more accurate dynamic model for prediction such as using an interaction model when predicting the position and velocity of a target [4], [5]; generating more stable recursive mathematical models such as unscented Kalman filters and MCMC particle filters [6]; and searching for more accurate measurement models [7], [8], for example, the trackinglearning-detection (TLD) [9] approach. However, the almost universally accepted mathematical framework used to describe multiple target tracking is that of filtering theory and, in particular, Bayesian filtering [10], where the posterior probability distribution is recursively predicted by propagating this distribution with the state model, which describes the motion of a target, and updates the state when a new observation becomes available [11].…”
Section: A Related Workmentioning
confidence: 99%
“…Most multiple human tracking approaches fall into one of three categories: achieving a more accurate dynamic model for prediction such as using an interaction model when predicting the position and velocity of a target [4], [5]; generating more stable recursive mathematical models such as unscented Kalman filters and MCMC particle filters [6]; and searching for more accurate measurement models [7], [8], for example, the trackinglearning-detection (TLD) [9] approach. However, the almost universally accepted mathematical framework used to describe multiple target tracking is that of filtering theory and, in particular, Bayesian filtering [10], where the posterior probability distribution is recursively predicted by propagating this distribution with the state model, which describes the motion of a target, and updates the state when a new observation becomes available [11].…”
Section: A Related Workmentioning
confidence: 99%
“…The mean shift always points toward the direction of gradient ascent [11] . At the end of the iterative algorithm, it is guaranteed to converge at a stationary point where the gradient of density estimate is zero [12] . By Eqs.…”
Section: Mean Shift Bias Estimationmentioning
confidence: 99%
“…Mean shift is an iterative process that shifts the data point t to the average sampling interval of data points [9] . The mean shift algorithm is a popular nonparametric clustering method [10] , and it has been applied to problems such as image segmentation [11] , target tracking [12][13][14][15] and fusion [16] .…”
Section: Introductionmentioning
confidence: 99%
“…As we initially have RGB images, the first thing we could do is calculate the statistical color distribution of object pixels, the color histogram, as in [1], [2] and many other tracking methods. However, the background may contain some object colors, and the object region may have some background color.…”
Section: Introductionmentioning
confidence: 99%