Related Experiment Video
Updated: Oct 2, 2025
06:38
Using a Virtual Reality Walking Simulator to Investigate Pedestrian Behavior
Published on: June 9, 2020
5.0K
Prediction of Pedestrian Crossing Behavior Based on Surveillance Video
Xiao Zhou1, Hongyu Ren1, Tingting Zhang1
1School of Mechanical and Electronic Engineering, Wuhan University of Technology, Wuhan 430070, China.
Sensors (Basel, Switzerland)
|February 26, 2022
Summary
Predicting pedestrian crossing behavior using surveillance video enhances autonomous driving safety. Our novel network integrates posture and context features for accurate predictions, outperforming existing methods.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Autonomous Systems
Background:
- Autonomous driving relies on accurate pedestrian crossing behavior prediction.
- Vehicle-mounted cameras have limitations due to blocked sightlines.
- Surveillance video offers supplementary data for critical road areas.
Purpose of the Study:
- To develop a novel network for pedestrian crossing behavior prediction using surveillance video.
- To improve the safety and reliability of autonomous driving systems.
- To explore the feasibility of using edge computing for real-time traffic safety.
Main Methods:
- Proposed a new pedestrian crossing behavior prediction network for surveillance video.
- Integrated pedestrian posture, local context, and global context features.
- Utilized a novel cross-stacked gated recurrent unit (GRU) structure.
Main Results:
- Achieved state-of-the-art results on the University of California, Berkeley surveillance video dataset.
- Demonstrated high accuracy and F1 score in pedestrian crossing behavior prediction.
- Analyzed the impact of prediction time and pedestrian speed on accuracy.
Conclusions:
- Pedestrian crossing behavior prediction using surveillance video is feasible and effective.
- The proposed network provides a viable supplementary system for autonomous driving.
- This research offers a reference for edge computing applications in autonomous driving safety.
