Real-time early warning method for non-motor vehicle conflicts based on multi-source perception

Abstract

The invention discloses a real-time early warning method for non-motor vehicle conflicts based on multi-source perception, including: acquiring multi-source data streams from road cameras and vehicle terminals through a sensor network; calibrating the time difference between data from different sources by using a timestamp synchronization module based on the multi-source data streams to obtain a time-aligned trajectory sequence; correcting the spatial deviation between acquisition devices by using a spatial coordinate conversion module from the time-aligned trajectory sequence to obtain a spatio-temporally consistent traffic participant position data set; obtaining a dynamically updated risk boundary path according to the traffic participant position data set; performing superposition analysis on the dynamically updated risk boundary path and complex environmental elements to determine the expanded risk area range; clustering the expanded risk area range by using a clustering module to obtain grouped traffic participants and generate an early warning signal distribution sequence. The invention may accurately identify potential conflicts in complex traffic environments.

Images