Pedestrian trajectory prediction method based on space-time attention mechanism

Abstract

A pedestrian trajectory prediction method based on the space-time attention mechanism comprises the following steps: acquiring image information in a scene, and extracting position information of a pedestrian in an image coordinate system; performing data preprocessing to obtain historical track coordinates of each pedestrian in the scene; by adopting an encoder Encoder, encoding the historical trajectory of the pedestrian, and outputting a feature tensor; and by adopting a decoder, iteratively predicting future trajectory coordinates of the pedestrian, wherein the encoder integrates the historical track information of each pedestrian and the interaction information among different pedestrians in the same scene through an attention mechanism. Real-time effective prediction of the pedestrian trajectory in an automatic driving practical application scene is realized, the method adapts to the processing capability of a vehicle-mounted low-power-consumption processor, the accuracy of pedestrian trajectory prediction is improved, a reliable basis is provided for practical automatic driving decision making, and the safety of automatic driving is greatly improved.

Description

Technical field [0001] The present disclosure relates to the field of automatic driving and computer visual technology, and more particularly to pedestrian trajectory prediction methods, devices, electronic devices, and storage media based on time and space attention mechanisms, and more particularly to a depth learning method under the interactive scene with complex pedestrians. Pedestrian trajectory prediction technology. Background technique [0002] With the development of computer vision technologies, environmental perceptions in computer vision have become an indispensable part of the automatic driving system and other intelligent perception systems. Among them, pedestrian trajectory prediction is of great significance in the field of automatic driving and video surveillance. In automatic driving scenes, predict the future trajectory of pedestrians, can assist automatic driving vehicles to make the right decisions, protect the safety and reliability of pedestrians and impro...

AI Technical Summary

Problems solved by technology

The mainstream methods generally use recurrent neural networks for time series prediction, including RNN, LSTM, GRU, etc. Typical methods such as Social-LSTM cannot be parallelized, and the calculation efficiency is low. It is difficult to model long-distance relationships, which is easy to cause For the technical problem of performance bottlenecks, this disclosure is based on the attention mechanism, which can effectively capture the key parts of the historical trajectories of pedestrians, and can guarantee performance in a parallelized style and global receptive field

Images