Multimode spatiotemporal fusion intelligent vehicle danger prediction method and system

Abstract

The application relates to the technical field of intelligent traffic and assisted travel, and discloses a multi-modal spatio-temporal fusion intelligent vehicle danger prediction method and system, which comprises the following steps: acquiring multi-modal sensing data around an intelligent vehicle, constructing a danger prediction model based on deep learning, the model comprising a spatio-temporal fusion module and a risk evolution module; the spatio-temporal fusion module aligns and fuses the multi-modal sensing data to obtain unified spatio-temporal voxel features, the risk attention gate module is used to focus the spatio-temporal alignment and fusion process on a high-uncertainty area, and the risk evolution module identifies objects around the intelligent vehicle according to the spatio-temporal voxel features, predicts the motion trajectory of the objects and potential collision risks; a world model is used to distill the danger prediction model, and the distilled danger prediction model is used for real-time danger judgment of the intelligent vehicle. The application can effectively fuse various sensing data to perform danger judgment during vehicle driving, and improve the judgment accuracy.

Images