The most difficult multiple target tracking problem includes multiple sensors with different viewing angles, measurement geometries, fields of view, accuracies, resolutions and scan rates. Such variations in sensor output characteristics as well as channel delays, countermeasures, inherent target features and maneuvers have solidified the consensus that an effective fusion system must handle several levels of “tracklets” from distributed sources in order to produce the desired long tracks as described in Waltz and Llinas (1990). In view of the increased attention given to hypersonics as well as the increased need for low-level signal processing, the computational complexity of track association is a vital factor in determining an autonomous vehicles’ ability to complete its objectives quickly. We are given a set of tracklets where the particular methods used to make the detections are taken for granted. Following joint probability density association filters, we assume short tracklets are completed (i.e, detections are correctly correlated with state estimates) and take a computational geometric approach to associating tracklets. If N is the number of short term tracklets, this method fuses them in O(N2). Using covariance as a distance, this report suggests the applicability of a class of sweep-line algorithms developed in computational geometry in data fusion.
Efficient Algorithms for Maximum Induced Matching Problem in Permutation and Trapezoid Graphs
