Visual Detection of Detectable Warning Materials by Pedestrians with Visual Impairments - Final Report

2.9 Models to Predict Visual Detection and High Conspicuity Ratings

We conducted several regression analyses on the data shown in Figure 40 and Figure 41 to determine if adding additional parameters would substantially increase the correspondence between the models and the data. Contrast (or log contrast), reflectance of the detectable warning, reflectance of the sidewalk, and a binary parameter which encoded patterned versus single-color detectable warnings were used. A second binary parameter was used in the models to encode colored (bright red, orange-red, federal yellow, pale yellow) versus achromatic (black, white, gray) detectable warnings. For these analyses, the brown detectable warning was included with the achromatic set. Note that for the patterned detectable warnings, reflectance of the white area was used in all analyses requiring a value for detectable warning reflectance. In a second series of analyses (not shown here) the reflectance of the black area of the patterns was defined as the value for detectable warning reflectance, but this did not substantially affect the coefficients for the model parameters. The three best-fitting and simplest models are described below.

The model to predict the percent of participants who could see the detectable warning from 8-feet (P_S8) is shown below. Its coefficient of determination is r² = .82.

P_S8 = 17* Ln(Contrast) + 20.3 + (7.3 if color is red or yellow).

The model to predict the percent of participants who could see the detectable warning from 26-feet (P_S26) is shown below. Its coefficient of determination is r² = .78.

P_S26 = 23.4* Ln(Contrast) - 26 + (9.6 if color is red or yellow).

The model to predict the percent of participants who judged the detectable warning to have high conspicuity (PHC) is shown below. Its coefficient of determination is r² = .82.

P_HC = .806* (Contrast) + 19* R_DW - 16.7 + (26.6 if color is red or yellow).

Where:

Contrast = percent luminance contrast (0-100)

R_DW = reflectance (0-1.0) of the detectable warning (for patterns, reflectance of white areas was used).