To help resolve the deficiency of existing target tracking algorithm in sensor networks, a GM(1,1) based target tracking algorithm (GMTT) is proposed in this paper. GMTT treats the target tracking issue in sensor networks as a state prediction problem of inadequate-information system and applies the gray theory to deal with this issue. It makes use of the GM(1,1) model to predict the region where the target may arrive in the next moment, and forwards tracking messages to this predicted region to detect and track the target with dense nodes. Simulations demonstrate that GMTT processes good merits in tracking probability, tracking precision and tracking delay.