Combinatorial Fusion Criteria for Real-Time Tracking

Abstract

We address the problem of automated video tracking of targets when targets undergo multiple mutual occlusions. Our

Combinatorial Fusion Criteria for Robot Mapping

“…Remark 1: Previous work using CFA ([8] [9], [11], [13], [19], [24], [30]) in various application areas have demonstrated that: (1) the combi nation of multiple scoring systems (features or evidence) would improve the prediction or classification accuracy rate only if (a) each of the scoring systems has a relatively good performance, and (b) the individual scoring systems are distinctive (or diversified), and (2) rank combinations perform better than score combinations under conditions (a) and (b) and other restrictions.…”

“…That is, we include d(f A ,f B ) as defined in formula (2) When plotting the graph of the rank/score function (hence it is called the rank/score graph) of scoring systems A and B on the same coordinate plane, the diversity measure can be easily visualized. Different diversity measurements have been considered in other application domains ( [1], [5]- [9], [12], [13], [19], [24], [30]). …”

“…The a pproach of the current proposal, following the practice of [9], [11], [13], [30], [30], is to also use the concept of the rank/score function to measure the diversity between A and B. That is, we include d(f A ,f B ) as defined in formula (2) When plotting the graph of the rank/score function (hence it is called the rank/score graph) of scoring systems A and B on the same coordinate plane, the diversity measure can be easily visualized.…”

Selection of fusion operations using rank-score diversity for robot mapping and localization

“…d(A,B) (dissimilarity or difference) between A and B has been studied previously using the score function diversity d(s A ,s B ), correlation, and rank function diversity d(r A ,r B ), rank correlation. Different diversity measurements have been considered in other application domains ([1], [5]- [8], [11], [12], [18], [22], [30]). A unique aspect of CFA is its definition of the rank-score function and associated diversity.…”

“…We evaluate a set of fusion operations of the stereo and sonar data with respect to ground truth information on depth. We also evaluate for each fusion operation the CFA criteria developed in [8] [12], namely, a feature performance ratio metric PR(A,B) for features A and B, and a feature rank-score diversity metric d(f A ,f B ). We show that, in the absence of any assumptions about the statistics of sensors or environments, or any calibration sampling, these two features can be used to predict when fusion will produce a more accurate depth measurement and when not.…”