Abstract: Traditional localization methods usually can not handle sensor networks which are affected by several practical factors, such as anisotropic deployment terrain, imprecise anchor node position and noisy range measurements. In this paper, a robust region localization approach is proposed to naturally address these factors by constructing a global constraint set. A projection method is given to compute feasible geographic region, which can assuredly bound actual node position. The infeasible holes within the region are also computed using non-convex constraints. To enhance the practicability of this method, an iterative and distributed implementation based on clustering. Simulation results show that the algorithm is not affected by deployment terrain and measurement noise, thus it is applicable for sensor networks.