A gaussian sum approach to the multi-target identification-tracking problem

A Bayesian approach to joint tracking and identification of geometric shapes in video sequences

“…In the past, tracking and identification have mostly been performed separately although there are a few exceptions [26,27]. While this is clearly suboptimal, it is partly due to the fact that different sensors have usually been dedicated to each task; e.g., radar for kinematic tracking, cameras for classification.…”

Practical system for tracking multiple maneuvering targets

“…In most real-time tracking, several adaptive algorithms based on statistical methods have been developed since 1970, [1][2][3][4] such as Singer's method, 5 the adjustable level process noise algorithm, 6 the input estimation, 7 the variable dimension filter, 8 the interacting multiple model ͑IMM͒ algorithm, 9 and the current statistical model and adaptive filtering ͑CSMAF͒ algorithm. 10 The association algorithm includes the nearest-neighbor association, 11 probabilistic data association ͑PDA͒ 12 the all-neighbor optimal filter, 13 the multiple hypotheses method, 14 integer programming, 15 the Gaussian sum approach, 16 joint probabilistic data association ͑JPDA͒, 17 practical PDA logic, 18 deep-first search-based data association, 19 Hopfield network based data association, 20 interacting multipattern probabilistic data association algorithm ͑IMPDA͒, 21 fast data association using multidimensional assignment, 22 and the integration of Bayes detection. 23 However, both the accurate state estimation and the practical data association remain in maneuvering targets in a dense clutter environment.…”

Rich probabilistic representations for bearing only decentralised data fusion

“…Gaussian mixture models are used as the probabilistic representation which was first considered in tracking problems by Alspach and Sorenson [9], [10]. Alspach also extended these ideas to multi-target identification [11].…”