Bayesian framework and fast Fourier transform based target tracking method

The invention discloses a Bayesian framework and fast Fourier transform based target tracking method. The method comprises the steps of: generating and normalizing a template kernel function; calculating a fast Fourier transform value of a target confidence map template; obtaining an initial position of a tracked target in a first frame image; obtaining a spatial context image of the target; calculating a point with maximum confidence in the current frame image; calculating a spatial context image of the target in a new position; updating a time-space context model and calculating a phase of Fourier transform of the time-space context model; and predicting a phase of a target spatial context image in a next frame image. According to the method, the template kernel function reduces the interference of a background image around the target on a target template; the resolution of the target context image is unified into a set pixel size and is combined with an application of fast Fourier transform, so that the timeliness of a target tracking method is improved; and the selection of the context image has relatively good tracking effects on the situations that the background change is little and the target is blocked to a certain extent.