The invention discloses a multi-feature-constrained network river mainstream identification method, which belongs to the field of map synthesis; mainstream identification is an important operation in the process of river system synthesis, and the accuracy of identification directly affects the quality of river system synthesis. In a river system with a large spatial range, there will be a network of rivers formed by the interweaving of river channels in some areas. The river sections in it are dense, complex in structure and similar in attributes, which greatly increases the difficulty of identifying the mainstream of the river system. The multi-feature-constrained network river mainstream identification method proposed by the present invention, firstly, builds the directed topology structure of river system data, and detects the key nodes that affect the mainstream identification, that is, redundant nodes; secondly, considers the flow direction of the river, and calculates each The effective topological boundary of redundant nodes is determined to determine its "influence domain", and a hierarchical tree is established accordingly; finally, considering the multi-feature constraints such as semantics, geometry, direction, topology, and hierarchical relationships, the optimal connection path between nodes is calculated to realize mainstream recognition.