A South Carolina Pilot Project Compared the Performance of Two Camera Technologies for Pedestrian Detection at Night to Prevent Crashes Between Vehicles and Pedestrians Due to Low Visibility.
Assessment of Safety Benefits of Technologies to Reduce Pedestrian Crossing Fatalities at Midblock Locations
South Carolina (SC) experiences high levels of pedestrian crashes and it was found that 80 percent of these crashes happened at night for the years 2007- 2016 due to the lack of illumination at crash locations. The objectives of this study were twofold: (i) characterize pedestrian nighttime crashes using the available database and provide short-term implementation solutions and (ii) to conduct a pilot test for comparing the effectiveness of camera technologies that can detect pedestrians under low light. Several data sources were used to understand the qualitative and quantitative aspects of pedestrian crashes, including the South Carolina Department of Transportation Crash Database, Crash Report Forms, Roadway Inventory Management Systems (RIMS) Database, and additional data extracted from online mapping tools. Midblock crashes were the focus of this study. The pilot test was to gauge the efficacies of two types of camera technologies – 1) Pan-tilt-zoom (PTZ) camera with night vision capabilities and 2) fixed focus thermal IP camera under four test conditions. The tested conditions included the following scenarios:
- Dark not lit
- Pedestrian detection with night vision camera in a dark not lit section
- Pedestrian detection with infrared camera in a dark not lit section
- Dark Lit
- Pedestrian detection with night vision camera in a dark section lit with vehicle light
- Pedestrian detection with infrared camera in a dark section lit with vehicle light
The tests were conducted at a roadway section where there were no streetlights. As pedestrian detection is relevant for a transportation system that consists of automated vehicles, there is great value in exploring opportunities to integrate technology that enhances driver and pedestrian capability and visibility.
- Choose thermal cameras for dark non-lit conditions. The study results indicated that for dark non-lit conditions thermal camera technology is effective in detecting pedestrians as they outperformed the night vision PTZ camera in every testing scenario. For all testing conditions, the infrared provides detectable pedestrian imagery up to 300 feet whereas the night vision camera only picked up discernible imagery for bio-motion up to 100 feet.
- Optimally place the night-vision cameras from the crossing location. The performance of the night vision camera deteriorated with the increasing distance of the camera from the crossing location. This also magnified the disparities in the headlight patterns with respect to pedestrian illumination in various positions in front of the vehicle.
- Employ practical and efficient lighting methods along with street networks. To improve conditions related to crossing crashes at night, the availability of street lighting is crucial along with pedestrian refuge islands on wide multi-lane facilities.
- Increase public awareness. Training and public service announcements that highlight the lack of pedestrian visibility and bring attention to higher risk sites are important to raise awareness among both drivers and pedestrians alike.
- Increase driver attention using other ITS technologies. Pavement flashing lights, advance stoplights, raised crosswalks and automatic pedestrian detection at signals are also some countermeasures that can be implemented to reduce pedestrian-vehicle crashes.