A particle filter track-before-detect based on local search sampling is proposed to deal with the low sampling
efficiency in high-dimension state space for the particle filter track-before-detect in a class of state partially observable
system. After the update of the posterior state, by using the prior distribution information of the unobservable components,
a kind of local search sampling strategy is executed around the estimate of observable components by a small amount of
particles, which improves the efficiency of state sampling for particles. Simulation results show that the new algorithm
obtains better detection and tracking performance.