基于知识学习的行人轨迹预测方法综述

Overview of pedestrian trajectory prediction methods based on knowledge learning

随着深度学习的快速发展,行人轨迹预测任务已经成为计算机视觉领域的研究热点之一,在自动驾驶、视频监控、人机交互等领域基于深度学习的行人轨迹预测方法得到了广泛应用。首先,介绍过去几年该领域的概况(特别关注基于知识学习的方法),将这些算法分成基于统计学模型的轨迹预测方法和基于知识学习的轨迹预测方法两大类,并分析每类方法的主要算法;然后,讨论行人轨迹预测任务中使用的数据集和常见的评估指标,对比基于知识学习分类的各个方法在主流数据集下的预测性能。最后,对行人轨迹预测的发展进行展望。

With the rapid development of depth learning,the pedestrian trajectory prediction method based on depth learning has a wide range of applications,including automatic driving,video surveillance,human-computer interaction.The purpose of this study is to provide an overview of the field in the past few years,with particular attention to the methods of deep learning.We divided these algorithms into statistical model based trajectory prediction methods and deep learning based over trajectory prediction methods,analyzed the main algorithms of each type of methods,discussed the data sets used in pedestrian trajectory prediction tasks and common evaluation indicators,compared the prediction performance of each method under the mainstream data sets,and finally looked forward to the development of pedestrian trajectory prediction tasks.