Pedestrian trajectory prediction method based on multiple hidden variable predictors and key points

Abstract

The invention relates to a pedestrian trajectory prediction method based on multiple hidden variable predictors and key points, and belongs to the technical field of computer vision and automatic driving. The method comprises the following steps: (1) processing a track sequence through an Informer-encoder to obtain a feature vector; (2) respectively generating a track hidden variable and a key point hidden variable through two different hidden variable predictors; (3) generating track key points through key point hidden variables, and initializing an Informer-decoder; and (4) combining the Informer-decoder with the track hidden variable to generate a prediction track, and finally calculating a loss function. According to the method, Informer is adopted as a basic network, two different hidden variable predictors are introduced to predict hidden variables of a trajectory and a key point respectively, the trajectory key point is predicted based on the hidden variables of the key point, a decoder is initialized by using the trajectory key point, the method can be used for trajectory prediction of an automatic driving vehicle for pedestrians, vehicles and other agents, helps the vehicle to make decisions better, and protects traffic safety.

Description

technical field [0001] The invention relates to a pedestrian trajectory prediction method based on a multi-hidden variable predictor and key points, and belongs to the technical fields of computer vision and automatic driving. Background technique [0002] The autonomous driving industry has also been hot for many years, but there is still no consensus in the industry on when it will truly realize unmanned driving. One of the most important reasons is that in complex scenes, it is difficult to make reasonable predictions about the trajectories of surrounding pedestrians. Therefore, if this problem cannot be overcome, no matter how perfect the decision-making and control technology is, it is impossible to achieve absolutely safe driverless driving. Therefore, pedestrian trajectory prediction has gradually become a hot research problem in the field of computer vision in recent years. As a vulnerable group, pedestrians most need protection from the outside world. In human dr...

AI Technical Summary

Problems solved by technology

Aiming at the problem of slow training and reasoning speed, the present invention adopts Informer's probabilistic sparse self-attention mechanism and generative decoder

Images