Determining the feasible set of motion states in target tracking using extended set-membership filter

This paper presents a new object recognition strategy for target tracking using extended set-membership filter (ESMF). Before the target recognition algorithms are applied on the sensory information, ESMF is employed to determine the demarcation of a guaranteed searching area in which the true position of the target can be certainly found. Then the target recognition algorithms will work in a smaller region rather than searching in the whole range of the sensory information. The new strategy can improve the speed and accuracy of the target recognition algorithms. Experiments are conducted to prove the feasibility of the strategy.