Traffic intersection is an important component of the road network. Accurate recognition of traffic intersections can lay a solid foundation for extracting road network information, planning driving paths, and navigating for vehicles. In this work, a multielement mathematical morphology filtering method is proposed to recognize and extract the traffic intersections from a map image. This method firstly constructs several different structure elements according to the image features of traffic intersections; then, each structure element is used to extract different features of the traffic intersections through erosion operation of mathematical morphology; next, the different features of the traffic intersections are fused by calculating the union of multiple erosion results; finally, based on the fused traffic intersection features, each traffic intersection can be recognized by connected domain labeling, and center coordinates of the traffic intersections are following obtained. The experimental results show that for the 211 traffic intersections in the map image, the proposed method can accurately recognize the locations of all the traffic intersections and reach 100% recognition rate, which can provide reliable input for subsequent path planning.