Node localization has been identified as a fundamental and essential issue in wireless sensor networks (WSNs), which is able to determine nodes' locations in location-aware applications and provide different types of services based on the location information. This paper presents a distributed approach based on local semidefinite programming (LSDP) to solve the localization problem in large-scale WSNs. At first, we propose a new technique to divide a large-scale WSN into several overlapped local networks, each of which creating a local map according to the LSDP algorithm. Then, a novel merging algorithm is presented to effectively merge all local maps into a global map, which contains the exact position of interested nodes. Extensive simulations demonstrate that our proposed method can achieve better scalability and higher accuracy and is also robust to the measurement noise. © 2014 IEEE.