针对串行粒子群优化(Particle Swarm Optimizer,PSO)算法存在计算量大、速度慢的问题,给出了一种基于数字信号处理(DSP)并行系统的并行PSO跟踪算法。在研制的4DSP并行系统上,采用基于消息传递模型及单种群的Master-Slave模式设计实现了并行PSO跟踪算法。用DSP-A实现初始化设置,其它3个DSP并行计算每个粒子的适应值。最后,由DSP-A比较每个粒子的适应值与其个体极值的优劣,选择较好的个体极值和整个种群的最优解,更新每个粒子的位置与速度。利用该系统采集实际序列图像进行了算法仿真验证,其加速比为2.525,效率为63.13％,该算法为全局优化大规模目标跟踪工程的实现提供了一个新的选择。

For the problem of a large amount and slow speed in the serial Particle Swarm Optimization(PSO) algorithm,a parallel PSO tracking algorithm based on Digital Signal Processing(DSP) parallel system is proposed. In the development of the four DSP parallel systems,a parallel PSO tracking algorithm is designed using the message passing model and the Master-Slave mode of a single species. The initial setting is realized by DSP-A,while DSP-B,DSP-C and DSP-D are used to calculate the fitness of each particle in parallel. Finally,the fitness of each particle and the pros and cons of individual extreme are compared by DSP-A,and then a better individual extreme and an optimal solution of the entire population are chosen to update the position and velocity of each particle. Comparing with the serial PSO algorithm,the speedup ratio and efficiency of the simulation algorithm based on the actual sequence of image are 2.525 and 63.13%,respectively. The method supplies a new option to implement a large-scale global optimization target tracking project.