Intelligent optimization algorithms are powerful tools to solve multi-dimensional nonlinear optimization problems and improve computational efficiency. A hierarchical artificial fish swarm algorithm is proposed in this paper, taking generalized subspace fitting constraint formula as the cost function. The proposed algorithm aims at the difficult engineering problem of huge computation in coherent radiation sources polarization-spatial angle joint estimation. The algorithm divides the fish swarm into a bottom layer and a top layer based on a hierarchical collaborative strategy. The bottom layer uses artificial fish swarms for global search to ensure population diversity. Meanwhile, the top layer adopts particle swarms for local search to speed up the convergence rate. The simulation results show that the hierarchical artificial fish swarm algorithm can greatly reduce the calculation amount of generalized subspace fitting, especially in the case of more targets. The algorithm can effectively improve computational efficiency and also provide better estimation accuracy than traditional artificial fish swarm algorithm.