Object Tracking Method Based on a New Multi-Feature Fusion Strategy

Jie Cao (Lanzhou University of Technology)
Lei-lei Guo (Lanzhou University of Technology)
Jin-hua Wang (Lanzhou University of Technology)
Di Wu (Lanzhou University of Technology)

Indonesian Journal of Electrical Engineering and Computer Science
Vol 12, No 9: September 2014

In order to solve the poor robustness problem due to partial occlusionof target, we propose an adaptive particle filter tracking method based on thefusion of the multiple features. First, we regard information fusion as aprocess of information loss, loss coefficient is defined based on this idea.Then the credibility of feature can be obtained by weighted distance and themean of particle filter, and we use the features’ credibility to reallocate theloss coefficient. An extensive number of comparative experiments show theproposed algorithm is more stable and robust than multiplicative fusion andadditive fusion tracking algorithms.