Traditional algorithm of indoor localization mainly relies on the Distance-Loss Model of RSS, which accuracy and stability are often poor. General algorithm based on RSS fingerprints database depends on the RSS value too much, so location accuracy affected by the density of reference node is bigger. Aiming at the shortcomings of the traditional algorithm, In this dissertation, we present the algorithm of indoor localization based on RSS and secondly fuzzy clustering. The secondly fuzzy clustering algorithm digs deeper mutual information, drops off the degree of dependence from the positioning accuracy to the density of reference node, and eliminates the influence of some noise point. Experiments show that the new algorithm, compared with the traditional positioning method based on RSS fingerprints database, improves positioning accuracy and stability.