A radar spatial registration method is proposed to overcome the problem of bias registration with uncorrelated multiple targets.The description of radar spatial biases and bias measurements with random finite sets(RFS) in earth-centered earth-fixed(ECEF) coordinates is derived, and the data association of multiple targets is avoided by using probability hypothesis density(PHD) filter.The spatial biases are recursive estimated and integrated by using the Gaussian mixture probability hypothesis density(GM-PHD) filter with uncorrelated multiple targets.Numerical results show that the proposed method outperforms the method with data association for radar spatial registration in precision and stability.