In order to decrease the running time, the number of iteration and effectively improve the reconstruction performance of Gradient Projection for Sparse reconstruction-Barzilai-Borwein algorithm, Particle Swarm Optimization which has the global search ability is introduced in it. Using PSO's global development ability and the local search ability of GPSR-BB algorithm, the convergence speed is increased and the running time is reduced. By the improvement of algorithm line search conditions, the reconstruction precision is improved effectively. Simulation results show that the improved GPSR-BB algorithm is shorter than the traditional algorithm by 43% in running time and by 39.7% in number of iteration. With the condition of a certain measurement dimension, the improved GPSR-BB algorithm is higher than the traditional one by 0.04 in average probability of success and lower than the traditional one by 0.09 in reconstruction error.