Appearance-based multiple hypothesis tracking

摘要: Soccer is a popular sport in the world with growth of demand for automatically analyzing matches and tactics. Since players are focus attention soccer videos they manage entire game, player tracking fundamental to most video analysis. An efficient implementation multiple hypothesis algorithm by evaluating its usefulness context introduced this paper. In contrast inherent linear assumption (MHT), which ignores appearance cues occlusions, our approach relies on an appearance-based MHT (AMHT) framework incorporating particle swarm optimization (PSO) account appearances, nonlinear movements occlusions. Experimental results demonstrate efficiency robustness algorithm. effective AMHT appearance, occlusions.A blob-guided PSO detection into MHT, capable detecting labeling resolving partial occlusions simultaneously. Only teammates occluding blobs detected erasing previously ones.Player solved within step.