Pedestrian trajectory prediction method and device, electronic equipment and computer program product

Abstract

The invention provides a pedestrian trajectory prediction method, which comprises the steps of obtaining observation trajectory information of at least one pedestrian in a scene, and converting the observation trajectory information of each pedestrian into self-view trajectory information under a self-view of each pedestrian; based on the self-view-angle track information of each pedestrian, obtaining the motion trend characteristics of each pedestrian, and obtaining the interaction characteristics of each pedestrian and other pedestrians; generating future position information of each pedestrian in a self-view angle at least based on the motion trend characteristics of each pedestrian and the interaction characteristics of each pedestrian and other pedestrians; and generating at least one future trajectory under the self-view angle of each pedestrian at least based on the future position information under the self-view angle of each pedestrian, and converting the future trajectory under the self-view angle into a future trajectory under a world coordinate system. The invention further provides a pedestrian trajectory prediction device, electronic equipment, a readable storage medium and a computer program product.

Description

technical field [0001] The present disclosure relates to the technical field of computer vision and the field of automatic driving. In particular, the present disclosure relates to a pedestrian trajectory prediction method, device, electronic device, storage medium and computer program product based on multi-viewpoint transformation. Background technique [0002] Trajectory prediction tasks are currently mainly used in autonomous driving tasks, and the trajectory prediction of other traffic participants in autonomous driving scenarios is of great significance for the realization of higher-level unmanned driving. The improvement of the performance of autonomous driving perception systems and the further development of deep learning on time series models have laid the foundation for the research on trajectory prediction tasks. The perception system obtains the historical location information of the target through various sensors and feeds it into the prediction model, and the...

AI Technical Summary

Problems solved by technology

In view of the above (2) problem of model complexity and too many parameters, this disclosure proposes to use a multi-head attention mechanism to extract trajectory sequence features, and use multiple attention heads to mine the representation of trajectory sequences in different hidden variable spaces. While improving the network performance, the network parameters are greatly reduced, and the cost of network training is reduced

Images