Predicting multiple target tracking performance for applications on video sequences

Juan E. Tapiero, Henry Medeiros, Robert H. Bishop
2017 Machine Vision and Applications  
This dissertation presents a framework to predict the performance of multiple target tracking (MTT) techniques. The framework is based on the mathematical descriptors of point processes, the probability generating functional (p.g.fl). It is shown that conceptually the p.g.fls of MTT techniques can be interpreted as a transform that can be marginalized to an expression that encodes all the information regarding the likelihood model as well as the underlying assumptions present in a given
more » ... technique. In order to use this approach for tracker performance prediction in video sequences, a framework that combines video quality assessment concepts and the marginalized transform is introduced. The multiple hypothesis tracker (MHT), Joint Probabilistic Data Association (JPDA), Markov Chain Monte Carlo (MCMC) data association, and the Probability Hypothesis Density filter (PHD) are used as a test cases. We introduce their transforms and perform a numerical comparison to predict their performance under identical conditions. We also introduce the concepts that present the base for estimation in general and for applications in computer vision. ii ACKNOWLEDGEMENTS Without the guidance, patience, and encouragement of countless people during my education, the completion of this dissertation would not have been possible. Thus, I would like to express my deepest thanks to those who have contributed to the process and helped me to reach this point. First, I would like to extend my gratitude to Dr. Robert Bishop for providing me with the opportunity to further my studies and pursue a doctorate degree. His guidance and feedback throughout the past few years has been invaluable to my research.
doi:10.1007/s00138-017-0840-8 fatcat:e3xxdfmqkrgdzngxgpelwemfzm