Intelligent multiple objects tracking techniques for location updates with occlusion handling in a dynamic environment

Abstract

This research thesis attempts to provide two signal processing based tracking mechanisms for single objects and two database based and two artificial intelligence based methods for multiple objects which have a dynamic background The Automatic Kalman Filter Bank proposed is based on signal processing technique that automates the filter selection KFB consists of three filters and is defined as an array of Linear Kalman Filter Extended Kalman Filter and Unscented Kalman Filter that separates the input newlinesignal into multiple components where each component is a sub section of the original signal and carries the prediction motion of the object The accuracy of filter selection is 98 3 The second signal processing based technique proposed is the Particle Grouping method that combines Kalman Filter and Particle Filter to track a single object The object detection is automated by pairing Haar training and discrete adaptive boost classifier In comparison with Particle Filter and Kalman Filter individually the combination of filters shows 30 increase in system performance newline newline