Human Factors Assessment of Pedestrian Roadway Crossing Behavior

Data collection validation

Just as in phase 1, the validity of the camera footage was assessed for the second set of cameras. Researchers manually scored pedestrian crossing behaviors at Location 10 (Connecticut and Van Ness Northwest), Location 13 (H and Connecticut Northwest), and Location 15 (Pennsylvania and 8th Southeast). Researchers recorded the pedestrian crossings and their interactions with vehicles over three 15-min periods at both locations. The in-vivo recordings were made the morning of March 1, 2012.

At Location 10, there was a 100-percent agreement in the classification of the pedestrian crossings for both the second and third 15-min segments. However, in the first 15-min session, there was a single discrepancy; the onsite coding resulted in 25 total pedestrian crossings and the video coding resulted in 24 crossings. At Location 13, there was a 100-percent agreement in the classification of the pedestrian crossings for both the second and last 15-min segments. However, in the first 15-min session, there was a discrepancy; the onsite coding resulted in 81 total pedestrian crossings and the video coding resulted in 75 crossings. It is likely that this difference is the result of the large number of people crossing during this short time frame. The video allowed the researchers to slow the rate of crossing for a theoretically more accurate count in crossings. At Location 15, there was a 100-percent agreement in the classification of the pedestrian crossings for both the second and last 15-min segments. However, in the first 15-min session, there was a discrepancy; the onsite coding resulted in 24 total pedestrian crossings and the video coding resulted in 22 crossings. Across all three sites, there was a 98.5-percent agreement in crossing classifications.

Video data coding

Many different types of pedestrian crossings and pedestrian interactions with vehicles in, and along, the roadway can be recorded. Although it is difficult to code pedestrian crossings in an exhaustive manner, the current study sought to record enough information to interpret general crossing behaviors. For each pedestrian crossing, multiple factors were recorded.

Crossing Factors

The same crossing factors as phase 1 were used to classify crossing behaviors in the second set of cameras.

Dates/Times Coded

Only sub-portions of the vast amount of video recorded data were coded owing to both time and project scope requirements of the current study. The scope of the present study included only daytime pedestrian crossings. As such, nighttime data were not examined.

Because there were no significant differences in pedestrian crossing behavior patterns based on the day of the week or the time of day in phase 1, fewer days and segments of video were coded in phase 2. Four days were randomly selected to code: February 24, February 27, March 1, and March 2. On each of these days, samples of crossings were selected throughout the day from 7:00 a.m. to 5:00 p.m., for a total of approximately 1.5 h of coded video per location, per day. The video feed from Location 11 on March 2 and from Location 13 on February 27 was blurred and could not be coded.

Weather

Table 18 describes the weather and corresponding sunrise/sunset times for each of the days that data were coded.

Table 18. Relevant sunrise, sunset, and weather for each of the coded data collection days.

— Indicates no recorded precipitation.