In order to solve the shortcoming that centroid algorithm is affected by the node distribution uniformity and a few anchor nodes will lead to large localization error in wireless sensor network, a circle centroid algorithm is proposed. In this algorithm, the unknown node communication region is divided into different circles which center is the unknown node itself, using the circle to exclude the anchor nodes which easy to increase the localization error and select the appropriate anchor nodes, the localization error will be further reduced if the approximate equilateral triangle is be found on circle. At the same time the method of using RSSI form circles is put forward. The simulation shows that the circle centroid algorithm compares with the centroid algorithm, location precision is improved by 11%, in the condition that 100 sensor nodes which radius is 20m are deployed randomly confined to the area of 100m×100m and the anchor node number is 20.