We present a nonlocal entanglement concentration scheme for reconstructing some maximally entangled multipartite states from partially entangled ones by exploiting cross-Kerr nonlinearities to distinguish the parity of two polarization photons. Compared with the entanglement concentration schemes based on two-particle collective unitary evolution, this scheme does not require the parties to know accurately information about the partially entangled states--i.e., their coefficients. Moreover, it does not require the parties to possess sophisticated single-photon detectors, which makes this protocol feasible with present techniques. By iteration of entanglement concentration processes, this scheme has a higher efficiency and yield than those with linear optical elements. All these advantages make this scheme more efficient and more convenient than others in practical applications. ; Comment: 7 pages, 4 figures. Physical Review A 77, 062325 (2008)