Node localization plays a critical role in wireless sensor network (WSN). The ill-posed problem caused by measurement error in the localization brings down the localization precision. Based on the Landweber iteration principle, a new multilateral localization algorithm is proposed, which includes the optimal reference node number, the iteration index selecting, termination rules and the accelerated Landweber iteration algorithm. Experiment results show that Landweber iteration algorithm obtains more satisfying performance than the well-known maximum likelihood estimation. The localization error is less than 1 m when the iteration index omega is 3.3times10-6. The optimal localization can be achieved about ten time's iteration when the termination rules of the accelerated Landweber iteration algorithm are sigmak les1.