Non-motorized transport is one of the most accessible and sustainable means of mobility. Shared pathways provide some of the highest level-of-service to cyclists and pedestrians and are important transport corridors. However, cycling and walking are two very different modes with unique characteristics. Obtaining information regarding these characteristics as well as the overall pathway mode demographics is essential to assessing and mitigating any real or perceived safety concerns that arise due to the combination of these modes. Presented here is an algorithm that obtains several shared pathway parameters automatically using computer vision techniques. This new algorithm classifies pedestrians and cyclists and calculates their positions and speeds based on real-world coordinates obtained from a perspective transform. The calibration required for the transform is very simple– only a single onsite measurement is needed. This simple algorithm allows for real-time detection and classification of moving objects. The algorithm has been implemented as part of the Pedestrian and Bicycle Tracking (PBTrack) system. This system is an extension of the Pedestrian Tracking (PedTrack) system developed at the Smart Transportation Applications and Research Laboratory (STAR Lab) of the University of Washington to greatly facilitate collection of pedestrian and cyclist information. Field experimentshave been conducted to test the accuracy of the algorithm. Preliminary results are presented and analyzed for mode split and speed distributions. Though the PB-Track system detected roughly 83% of the target objects and detection error for each mode is significant, the mode split ratios attained by the system are within 2.6% of the true value. The PB-Track system hasthe potential to be a valuable tool for shared pathway planning and analysis.
Samenvatting