Road User Abnormal Trajectory Detection using a Deep Autoencoder

EasyChair Preprint 485

Abstract

In this paper, we focus on the development of a method that detects abnormal trajectories of road users at traffic intersections. The main difficulty with this is the fact that there are very few abnormal data and the normal ones are insufficient for the training of any kinds of machine learning model. To tackle these problems, we proposed the solution of using a deep autoencoder network trained solely through augmented data considered as normal. By generating artificial abnormal trajectories, our method is tested on four different outdoor urban users scenes and performs better compared to some classical outlier detection methods.

Keyphrases: Abnormal trajectory detection, abnormal data, abnormal event, abnormal event detection, abnormal trajectory, anomaly detection, data augmentation, data augmentation technique, deep autoencoder, machine learning, normal abnormal, normal data, normal trajectory, realistic abnormal trajectory, road user, trajectory data, trajectory sample, unsupervised learning, user abnormal trajectory detection

@booklet{EasyChair:485,
  author    = {Pankaj Roy and Guillaume-Alexandre Bilodeau},
  title     = {Road User Abnormal Trajectory Detection using a Deep Autoencoder},
  doi       = {10.29007/xwfw},
  howpublished = {EasyChair Preprint 485},
  year      = {EasyChair, 2018}}