Particle Filter Algorithm Based On Cluster Intelligence In The Direction Of Target Tracking

Target tracking technology in the daily application is more and more extensive, and is close to people’s life. It is widely used for navigation, intelligent video surveillance, stereo camera, close sensors, etc. As people to target tracking accuracy requirement is higher, many scholars began to research the new target tracking algorithm or improvement of the existing target tracking algorithm. The particle filter algorithm is the mainstream algorithm of target tracking. Improving the particle filter algorithm proposed to obtain better tracking performance is of great significance. The following are the main work of this paper:1. This paper introduces the three aspects of target tracking theory which are target modeling, object recognition and detection, target filter and forecast.2. This paper analyzes the similarity of the genetic algorithm, particle swarm algorithm, and artificial fish school algorithm with particle filter algorithm, given the possibility of integration of the above algorithms with the particle filter algorithm. Combined with the target tracking, the strategy of the evolution thought ofthe genetic algorithm, the idea of particle swarm optimization and the thought of artificial fish school algorithm are introduced into the particle filter algorithm. The simulation results show that particle distribution of the improved algorithm is more close to true the posterior distribution than standard particle filter algorithm. The improved algorithm increases the diversity of the particle, and overcomes the disadvantages that particles easily lose diversity.3. Genetic algorithm has characteristics of the fast convergence, the strong global search ability. It makes the whole solution move evenly to the optimal solution area, but with the algorithm running, the convergence will slower. Artificial fish school algorithm and particle swarm algorithm have strong local search ability, but the algorithm convergence is slow, according to these above characteristics, I put forward two kinds of the improved algorithm. One is genetic particle swarm particle filter algorithm and the other is genetic artificial fish particle filter algorithm. The simulation results showed that particle distribution of the hybrid cluster particle filter (genetic particle swarm, genetic artificial fish) is close to true the posterior distribution than artificial fish particle filter algorithm and particle swarm particle filter algorithm. It improves the prediction and estimation precision of target tracking.