The invention discloses a visual tracking method based on target characteristics and bayes
filtration. The method comprises the following steps: a
system model and an observation model are established according to the actual motion of a target; the color and the gradient of the target are calculated, similarity function is constructed, and the current observed value of the target is obtained by a
particle filter; the state average of particulate matter and the
covariance are processed by using karman
filtration, thus generating new gauss distribution, then new particulate matter is sampled according the gauss distribution generated, thus calculating weight and output; finally, the particulate matter is sampled again; meanwhile, a partition detection method for the target and the corresponding
processing algorithm of shading and non shading are proposed; the visual tracking process is finished. Compared with similar
algorithm, the method realizes information
complementation between the characteristics by the blending of multi-information, therefore, the target is not easy to be affected by external environmental factors such as light, etc. By adopting the combination of particle
filtration and the karman filtration, the tracking accuracy of the whole method is higher, thus improving the tracking performance and being adapted to various complicated environments.