Visual Performance, Comfort and Pleasantness for Pedestrians, Cyclists and Residents - Road Lighting: Fundamentals, Technology and Application - page 62

Fig. 5.3 Facial identification
distance versus
semi-cylindrical illuminance
at face height (light source
used: white high-pressure
mercury light); TI between 8
and 15 %. (Van Bommel and
Caminada 1982 )

15

d face
(m)

10

5

0
0 1 2 3 4

E semi-cyl (lux)

for people in the street who do not know each other. From Fig. 5.3 we can see that for
this a minimum semi-cylindrical illuminance of about 1 lx is required. Later studies
on proximity zones show that interpersonal distances vary with lighting level, indoor
or outdoor environment, and especially with the experimental procedures employed
(Fotios and Yang 2013 ).
Since the first investigation on facial recognition under road-lighting conditions,
many more investigations have been carried out by different researchers. One of the
aspects investigated is whether or not the spectrum of the light has an influence on
facial recognition. The results of these studies are conflicting. Many show an effect
of the spectrum in that light sources with better colour rendering facilitate facial
recognition (Raynham and Saksvikrønnin 2003 ; Knight et al. 2007 ; Qi Yao et al.
2009 ). However, another, extensive, study does not show any effect of the spectrum
on facial recognition (Alferdinck and Hogervorst 2013 ). The same holds for a study
by Rea et al. ( 2009 ). Lin and Fotios ( 2013 ) have investigated what the reasons can
be for these conflicting results. A probable cause is differences in task difficulty.
When the task is relatively easy in terms of large visual size (i.e. short observation
distance) and long observation time, the spectrum has no influence. However, in
the case of small tasks (i.e. larger observation distances) and shorter observation
times, the spectrum does have a significant influence. They also conclude that in
the case of difficult tasks, where the spectrum does have an influence, that influence
cannot be directly related to one of the familiar spectrum characteristics, such as
colour rendering index (Ra), colour temperature (T k ) or S/P ratio. Further analysis
is needed to establish a spectrum metric that relates to facial recognition.
Again other studies concentrate not so much on facial recognition but on the
possibility to determine the emotion expressed by the face (Montel and Menar 1989 ;
Fotios and Yang 2013 ). A Japanese study (Kohko et al. 2008 ) has determined in
some detail the influence of glare (expressed by the value of the threshold increment
TI) on facial recognition. Some of these studies are not expressed in terms of semi-
cylindrical illuminance but in terms of facial luminance, which makes the studies