A fast visual tracking algorithm based on circle pixels matching
Zhiqiang Hou, Chongzhao Han, Lin Zheng · 2003
A fast visual tracking algorithm based on circle pixels matching is presented in this paper; this algorithm fused the ideas of correlation-based tracking method, serial similar detection algorithm and feature-based tracking method. The Circle pixels are adopted as feature pixels set, so it doesn't have high calculation cost to extract the feature pixels. Based on using the idea of serial similar detection algorithm, this algorithm adopts coarse- to-fine strategv to search the correct target position, coarse stage is used to decide the preparative matching windows andfine stage is used to decide the final correct window. Finally, normalize cross-correlation function is used to decide the template change strategy. Simulations show that this algorithm not only could track stably, but also the computational cost is greatly decreased. Compared with correlation-based tracking method and serial similar detection algorithm, this algorithm has superiority in calculation cost.