A Combined Two-Stage Local-Spatial Interest Point Matching Algorithm

local-spatial interest point matching algorithmfor articulated human upper body tracking application isproposed in this paper. The first stage finds confidently matchedpairs of interest points from the reference and target interestpoint lists through a local-feature-descriptors-based matchingmethod. Applying two cross-checking and displacement-checkingsteps reduces the number of mismatched pairs and resultsconfidently matched pairs. Using these confidently matchedpairs, the second stage recovers more matched interest pointpairs from the remaining unmatched through the graphmatching by a cyclic string matching algorithm. The proposedapproach benefits from the speed of local matching algorithms aswell as the accuracy and robustness of spatial matching methods.In addition, it compensates for the reference list leakage problem.Experimental results show that the combined two-stage interestmatching method efficiently improves the matching process forarticulated human upper body tracking.

کلیدواژه ها:

interest points matching ، cyclic string matching ، human body tracking