Exploring Crowdsourced Monitoring Data for Safety - Evaluation of Miovision Pedestrian Count Data (TTI-Student-05)

Project Description: The data represent one week of selected hourly weekday and weekend pedestrian counts at two intersections in Austin, Texas. The pedestrian counts were produced from manually reducing video files from Miovision’s TrafficLink Multimodal Detection and Counts system. Eighty hours of video were gathered at each intersection between June 18 and July 14, 2019. However, only 40 hours at each intersection were reduced and evaluated. The manual counts were compared to Miovision’s count data across different combinations of lighting conditions and pedestrian volumes. Overall, Miovision system performed fairly well with accuracy results of 15% error for daytime and 24% for nighttime for the combined intersection legs. Data Scope: Selected hourly weekday and weekend pedestrian counts. 2 unique count locations. 240 hourly observations from 2 student workers and the Miovision system. Data Specification: Intersection - text; Cameron Cross, MLK at Guadalupe; no missing values Leg - text; North, South, West, East; no missing values Hour – numeric; 1-40; no missing values Date – date stamp; no missing values Time - Timestamp; no missing values Student 1 - numeric; 0-99; missing values Student 2 - numeric; 0-90; missing values Benchmark - numeric; 0-95; missing values Miovision - numeric; 0-134; no missing values