DSpace at IIT Bombay >
IITB Publications >
Please use this identifier to cite or link to this item:
|Title: ||Tracking of multiple-point targets using multiple-model-based particle filtering in infrared image sequence|
|Authors: ||ZAVERI, MA|
|Keywords: ||monte-carlo methods|
|Issue Date: ||2006|
|Publisher: ||SPIE-INT SOCIETY OPTICAL ENGINEERING|
|Citation: ||OPTICAL ENGINEERING, 45(5), -|
|Abstract: ||Particle filtering is investigated extensively due to its importance in target tracking for nonlinear and non-Gaussian models. A particle filter can track an arbitrary trajectory only if the target dynamics models are known and the time instant when trajectory switches from one model to another model is known a priori. In real applications, it is unlikely to meet both these conditions. We propose a novel method that overcomes the lack of this knowledge. In the proposed method, an interacting multiple-model-based approach is exploited along with particle filtering. Moreover, we automate the model selection process for tracking an arbitrary trajectory. In the proposed approach, a priori information about the exact model that a target may follow is not required. Another problem with multiple trajectory tracking using a particle filter is data association, namely, observation to track fusion. For data association, we use three methods. In the first case, an implicit observation to track assignment is performed using a nearest neighbor (NN) method for data association; this is fast and easy to implement. In the second method, the uncertainty about the origin of an observation is overcome by using a centroid of measurements to evaluate weights for particles as well as to calculate the likelihood of a model. In the third method, a Markov random field (MRF)-based method is used. The MRF method enables us to exploit the neighborhood concept for data association, i.e., the association of a measurement influences an association of its neighboring measurement. (C) 2006 Society of Photo-Optical Instrumentation Engineers.|
|Appears in Collections:||Article|
Files in This Item:
There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.