Traffic sensors for detection of pedestrians in crosswalks have not been sufficiently robust. Either false positives indicating the presence of pedestrians when none are there or false negatives indicating the absence of pedestrians when pedestrians are present are issues with all technologies. This project will extend the capabilities of vision-based pedestrian sensors by adding a stereo vision component. A modification is pending, which will add an additional $500,000 for additional field testing. The project will: Verify whether stereo vision performs better than monocular vision.Identify critical gaps regarding the collection and processing of stereo vision data and what hardware changes or algorithms might be used to enhance stereo pedestrian detection techniques.Advance laboratory and field tests of improved hardware and algorithms as well as field operational tests. This falls under the following spreadsheet subcategory:Develop improved sensors and sensing technologies to support transportation applications and performance measures.
Establish the practicality of whether, in the real world, it is possible to use a stereo imaging sensor to detect pedestrians reliably.