V2I-BEVF: Multi-modal Fusion Based on BEV Representation for Vehicle-Infrastructure Perception | Synapse