There are a large number of false intersection points when using multiple passive observation stations for making bearing-crossing localization to multiple targets.To solve the problem, a new algorithm is proposed, which is based on improved K-means cluster and data fusion.The intersection points on the same direction line are clustered, so that the targets in the direction line can be estimated.In this way, the most of false intersection points are eliminated.And then the results of clustering obtained from different stations are fused step by step.In the process of fusion, the algorithm makes full use of the clustering result of each observation station to reduce the influence of the residual false intersection points.Therefore,more accurate results of target location are obtained.The result of simulation show that the algorithm has better locating performance and robustness in multi-station cross positioning.