Prediction Based Object Tracking Algorithm Using Kalman Filtering

Abstract

Object tracking have been broadly useful to robotics control, video retrieval, traffic surveillance and homing
technologies. A bundle of object tracking algorithms have been reported in literatures, but the region is still
incomplete with an efficient algorithm which cannot only track the objects but at the same time capable to identify the
direction and movement of object. In this, an efficient object tracking system is proposed based on Modified mean
shift tracking (MMST) algorithm. In this project essentially deal by how to tackle the problem to estimation the scale
and also direction change of the target over the mean shift tracking framework. In the original mean shift tracking
algorithm, the location of the target be able to well estimated, whereas the scale and direction changes cannot be
adaptively estimated. Consider that the weight image originated by the target model and the candidate model can
characterize the opportunity that pixel belong to the target, in this project work it will illustrate that the original mean
shift tracking algorithm can be derived by means of the zeroth and the first order moment of the weight image. With
the zero moment and the Bhattacharyya coefficient among the target model and candidate model, a easy and efficient
method is projected to estimate the range of target. Subsequently an approach, which utilizes the approximate area
and the second order centre moment, is planned to adaptively estimate the width, height and direction changes of the
target.