Rapid technological advances have enabled the development of wireless sensor networks for various monitoring tasks. Usually the involved applications are dependant on the location knowledge of the low-cost sensor nodes, the majority of which are non-beacon nodes whose positions are yet to be discovered. We present Argus, a light-weighted position estimation scheme for sensor networks, to address the problem of localizing non-beacon nodes, with the location references obtained from a few beacon nodes whose positions are known apriori. We first determine certain geometric reference points and evaluate them with a voting procedure. Then the position of a non-beacon node of concern is estimated with the geometric centroid of the identified most valuable reference points. Simulation results show that even when a few of the available beacon nodes are malfunctioning, our scheme can tolerate those misleading location references, and still provide a dependable localization service.