A cross-domain roadside perception multi-vehicle association method and system

Abstract

The present invention relates to a method and system for associating multiple vehicles in a cross-domain roadside perception, the method comprising the following steps: constructing a network graph structure for each roadside sensor node in a region, for integrating sensor perception data and calculating the corresponding node traffic status; constructing a spatiotemporal transfer model of traffic status based on a dynamic graph, and calculating the spatiotemporal correlation similarity between trajectory pairs under each node based on the model; extracting vehicle appearance features, fusing them with the spatiotemporal correlation similarity of trajectories to obtain an enhanced similarity index, and calculating the normalized enhanced similarity index using a spatiotemporal feature mask; constructing a similarity matrix for vehicle perception trajectory data sets of different nodes, and realizing the corresponding vehicle target association based on the similarity matrix. Compared with the prior art, the present invention is oriented to a roadside sensor network at the scale of an urban road network, and can effectively realize the rapid and accurate association of roadside perception data by comprehensively utilizing vehicle appearance features and the spatiotemporal transfer characteristics of traffic status between sensor nodes.

Images