Aiming at relative inaccuracy in positioning accuracy of traditional DV-Hop (Distance Vector-Hop) localization algorithm, the improved DV-Hop localization algorithm based on curve fitting is proposed. The improved algorithm regards the error between the actual straight-line distance and the estimated distance as one of samples, and combines the corresponding hops to search the relationship between the hops and the corresponding distance error based on curving fitting. The results of curve fitting are employed to estimate and forecast the distance error between unknown nodes and the anchor nodes, then to correct the estimated distance and decrease the positioning error. The simulation results show that the improved DV-Hop algorithm can preferably estimate the vital distance, and significantly improve the positioning accuracy.