Action and intention recognition of pedestrians in urban traffic

TL;DR

This paper explores various feature extraction and machine learning methods to recognize pedestrian actions and intentions in urban traffic, focusing on head orientation and motion to improve autonomous vehicle safety.

Contribution

It evaluates different feature extraction techniques combined with machine learning algorithms for pedestrian action and intention recognition using the JAAD dataset.

Findings

Achieved 72% accuracy in head orientation estimation

Achieved 85% accuracy in motion detection

Provides insights into effective features for pedestrian behavior prediction

Abstract

Action and intention recognition of pedestrians in urban settings are challenging problems for Advanced Driver Assistance Systems as well as future autonomous vehicles to maintain smooth and safe traffic. This work investigates a number of feature extraction methods in combination with several machine learning algorithms to build knowledge on how to automatically detect the action and intention of pedestrians in urban traffic. We focus on the motion and head orientation to predict whether the pedestrian is about to cross the street or not. The work is based on the Joint Attention for Autonomous Driving (JAAD) dataset, which contains 346 videoclips of various traffic scenarios captured with cameras mounted in the windshield of a car. An accuracy of 72% for head orientation estimation and 85% for motion detection is obtained in our experiments.