An efficient implementation and evaluation of Reid's multiple hypothesis tracking algorithm for visual tracking

Semantic world modeling using probabilistic multiple hypothesis anchoring

“…This is an approximation since there is no guarantee that the actual solution is included in the set of n hyp -best hypotheses. To further reduce the computational and storage overheads, we use track trees as in [26]. In the track trees all anchors are stored and the hypotheses contain pointers to the leafs of the track tree rather than the actual anchors.…”

“…The probability of the association set θ k given a parent hypothesis Θ k−1 p(h) is, inspired by [26]:…”

“…To avoid a brute force implementation that enumerates all legal associations, we have, like in [26], implemented a strategy that generates the n hyp most probable hypotheses only. This is an approximation since there is no guarantee that the actual solution is included in the set of n hyp -best hypotheses.…”

Multiple Hypothesis Tracking in camera networks

“…These hypotheses will eventually be pruned, after the hypotheses for all leaves are generated, but the processing time and memory space that the explicit enumeration of all these hypotheses consumes is insupportable. A solution is to use an algorithm due to Murty to find the ranked k-best assignments for the association in each leaf [10], instead of explicitly enumerating all the possible hypotheses. Clustering, which consists of dividing the hypothesis tree into several trees taking advantage of the independence between the tracks of some targets, can also be used to reduce the processing requirements of MHT and increase its performance [21].…”

Untitled

“…The classical target tracking literature provides a number of methods for data-association (Bar-Shalom & Fortmann, 1988;Popoli & Blackman, 1999) that are used in computer vision (Cox, 1993) and CML (Cox & Leonard, 1994;Feder, Leonard, & Smith, 1999), such as the track splitting filter (Zhang & Faugeras, 1992), the Joint Probabilistic Data Association Filter (JPDAF) (Rasmussen & Hager, 1998), and the multiple hypothesis tracker (MHT) (Reid, 1979;Cox & Leonard, 1994;Cox & Hingorani, 1994). Unfortunately the latter, more powerful methods have exponential complexity so suboptimal approximations are used in practice.…”