Vehicle-Pedestrian Conflict Scenarios Easily Missed by Pedestrian Collision Warning System

Vehicles equipped with pedestrian collision warning (PCW) systems can alert drivers of possible collisions in advance. Analyzing features of cases that are missed by PCW is an effective way to improve these systems. Existing evaluation of these systems using field tests or simulations, however, may fail to reflect real-world traffic condition. In addition, they seldom consider the complex mixed traffic environment. In this paper, the dataset of the Shanghai Naturalistic Driving Study (SH-NDS) was used, in which vehicles were equipped with Mobileye ® systems with PCW function. A motion profile-based approach was adopted to extract 272 vehicle-pedestrian conflict events from front-camera videos. PCW alert states were examined for each event, and those triggering PCW were classified as warned events while the remainder were missed events. Detailed events information, including conflict type and pedestrian movement direction, was then obtained and coded for all the extracted events. These features were compared between warned and missed events, in order to discover the characteristics of scenarios that Mobileye's PCW often misses. Results show that the followings are typical features of missed scenarios: pedestrians crossing in the right-hand half of the driver's vision when the subject vehicle is about to finish a left-turning maneuver; pedestrians entering from two sides of the vision field; pedestrians obstructed by buildings or vehicles; pedestrians with suitcases, opening umbrellas, bags and barrows/prams; and subject vehicle with mean speed under 6 km/h, between 18 km/h and 24 km/h and over 30 km/h. These findings can be used to improve PCW systems.

  • Supplemental Notes:
    • This paper was sponsored by TRB committee AND30 Standing Committee on Simulation and Measurement of Vehicle and Operator Performance.
  • Corporate Authors:

    Transportation Research Board

    ,    
  • Authors:
    • Zhou, Qingya
    • Wang, Xuesong
  • Conference:
  • Date: 2019

Language

  • English

Media Info

  • Media Type: Digital/other
  • Features: Photos; References;
  • Pagination: 4p

Subject/Index Terms

Filing Info

  • Accession Number: 01697896
  • Record Type: Publication
  • Report/Paper Numbers: 19-05610
  • Files: TRIS, TRB, ATRI
  • Created Date: Dec 7 2018 9:41AM