An algorithm of particle swarm optimization (PSO) provided with a dead zone initialization was put forward, according to the problem of PSO in the local optimum, research on particle swarm optimization. The aim was to breakthrough the limitation of local convergence, according to the observation on the MATLAB visual particle trajectories. It analysed the characteristics of particles trapped in local optimum, and run appear stagnation phenomenon in the process of particle swarm, painting for the center of the current local optimal particle "dead-zones", and initializing "dead-zones" in the particle. The results of simulation show that the optimization precision of PSO algorithm are improved, making use of the optimal standard of test functions of the algorithm improved.