Abstract: Objective Map patterns in building groups,as one of the essential foundations for cartographic gener-alization and multi-scale connection-relations,embodies the relationship of the material form of citiesto their social-economic functions.On the basis of related research at home and abroad,a multi-con-nected linear pattern is recognized taking advantage of parameter discrimination.With the analysis ofpattern organization laws,the structural parameters of linear patterns are characterized by distance,direction and size;then the neighborhood relationship was captured by proximity graph with the helpof a Delaunay triangulation;finally the multi-connected linear pattern is recognized by pruning theproximity graph,modeling the human processes.Experiments show that this approach is effective,feasible and practicable for multi-connected linear pattern recognition in agreement with cognitivecharacteristics.