Research Papers

Enabling Automated Pedestrian Data Collection and Safety Analysis in Urban Intersections

Version 1
Date added June 17, 2014
Downloaded 0 times/fois
Category 2014 CMRSC XXIV Vancouver
Tags Research and Evaluation, Session 4B
Author/Auteur Tarek Sayed, Mohamed H. Zaki
Stream/Volet Research and Evaluation

For Slideshow Presentation please contact Dr. Tarek Sayed (tsayed@civil.ubc.ca)(no paper submitted)

Abstract

This paper demonstrates the potential of using computer vision techniques for solving several shortcomings associated with traditional pedestrian safety analysis and data collection. In the first application, automated pedestrian data collection in terms of average speed and counts is demonstrated with high accuracy. Crossing speed is an important design parameter as it determines the time required for safe pedestrian crossing at the intersection. The lack of reliable pedestrian count data can be challenging to justify capital for implementing pedestrian facilities. This application shows also the potential of automated gait analysis for the pedestrian attributes identification (age and gender). The demonstrated case study reported correct classification rates of 80%. The second application in this paper demonstrates an automated proactive safety diagnosis for pedestrian traffic. Traffic conflicts are suggested as surrogate data for pedestrian safety. Traffic conflicts provide invaluable information that can be used to better understand collision contributing factors and the collision failure mechanism. Automated traffic conflicts technique is applied and validated on several data sets.  Also, pedestrian compliance to traffic regulations is automatically analyzed. Automated spatial violation detection (jaywalking) is demonstrated with accuracy greater than 90% is reported on the data sets.  The final application addressed in this paper is the automated classification of road-users. A classification approach relying on the movement characteristics of the road users is proposed. The application of the classifier has shown a correct classification rate of around 90%.  The applications are applied on several locations in Greater Vancouver in addition to other international cities. Overall, the three applications demonstrate the considerable potential of using video-based computer vision techniques for automated pedestrian safety analysis and data collection. This line of research benefits safety experts as it provides a prompt and objective safety evaluation for intersections. It also provides a permanent database for traffic information that can be beneficial for a sound safety diagnosis as well as for developing safety countermeasures.

Tarek Sayed, Mohamed H. Zaki