This work considers a system of visual tracking of football players, which esteem the trajectories 2D in sequences of images captured for fixed cameras, installed around of a soccer field. The tracking allows to determine diverse parameters as, for example, the speed and in the distance covered by the players during a departure, the region of bigger movement of the team in the field, etc. The considered system follows a probabilist boarding, reaching good initial results. The experiments had shown that an initial implementation of system, only using information of variation of the luminous intensity of the images, already presents resulted efficient and efficient and allows to glimpse significant improvements to be developed.