Volker Schmidt wrote:
>I am very cautious about any of this kind of measurement for the following
>reasons:
>1) the results will be very difficult to standardise
>2) the effort is far beyond that what a mapper can reasonably do.
Oh well, I guess I'll have to write a comment here, because I recently finished
my master's thesis on a related subject in street lighting research. While I
wrote this message in parts on several days, there might be some repetition in
it, but I hope the ideas are comprehensible.
As a background, the eye requires a constrast between the target and the
background before the target can be seen; the contrast can be in the colour, or
in the luminance, or both. The eye also adapts to the prevailing luminance
level; there's no exact model to predict the "adaptation luminance" given any
scene, but the models of human vision take the adaptation level as a starting
point - most scientific experiments have used a constant luminance background
and a sufficiently long adaptation period (5+ or 20+ minutes) to fixate the
adaptation luminance.
The road planners have several lighting classes, which apply to different types
of roads and pedestrian environments. The classes are not globally identical,
but the basic ideas behind those classifications should be roughly similar.
Generally the lighting classes set minimum requirements for the average
illuminance or luminance on the ground, and a requirement for the evenness of
the measured value, and possibly limits for measured glare. Sometimes there is
a requirement that in the area next to the road the luminance should be a
percentage of the value measured on the road. Some countries require that
pedestrian environments fulfill some luminance condition on vertical surfaces,
too, and some lighting classes might require or favour sufficient colour
reproduction. These are the measurable quantities, and they are quite well
predicted already in the planning phaze.
When the lights get older and dirtier, less light reaches the road surface, so
the new installations typically exceed the requirements. Lighting installations
might be up to 40 years old, but some have been replaced earlier. The expected
lifetime is often 15 to 20 years in the planner's operating cost calculations.
In practice (assuming conditions found on normal roads, i.e. normal
cost-optimized installations) the amount of light and the lack of glare are the
most effective predictors for the average assessed quality of lighting. On ways
used for vehicular traffic, glare seldom is an issue (but for example some
early LED lights had a glare problem).
There have been numerous studies, and they have compared the users' assessments
to other attributes. When the test subjects are pedestrians, things like
perceived openness of the area, emphasising the natural elements in the area
with the lighting, perceived (lack of) options for escape and the ability to
recognize other people's faces/intentions correlate with "better" lighting -
and lighting can improve users' perceptions of those "nonlighting" attributes.
Nobody has proposed a concrete model which could predict the "perceived
quality" with all the recognized parameters. Such a model would require, for
example, knowledge of the local living conditions and people's expectations of
personal safety: there's a huge difference in what primes people into fear,
between crime ridden environments and countries where street crime is very low.
Measuring the road luminances is standard practice. They used to have to
position the measurement device at regular intervals for measurements, but
nowadays they use calibrated digital cameras with special software and do the
measurements for a stretch of road surface from one picture. The "officially
acceptable" devices cost more than your average DSLR camera, but from what I've
read, the results could be sufficiently accurate for this kind of tagging.
The problem is then that the road has to be empty, the tail and headlights of
other vehicles would distort the values, and that to get comparable results the
street has to be dry and the height needs to be constant; the road surface
isn't a totally diffuse reflector (and wet surface even less so) so the values
depend somewhat on the angle between the viewing direction and the road
surface. The measurement "grid" has to be manually positioned over the picture,
to get a standard sample between and of the whole area between two light poles.
If, on the other hand, one were to measure "upward, i.e. the mobile device
measured the amount of light reaching it's light sensor and not the luminance
of the surfaces visible in the camera, there are other hindrances. The sensor
basically integrates over the half sphere space angle (or a smaller aperture),
and the user holding the mobile phone blocks a significant portion of that; the
old method for road lighting measurements had the persons doing the job walk
away from the sensor before noting down the reading.
In essence, the luminance/illuminance measurements could be crowdsourced, but
to get consistent numbers, there's no start-and-forget method, even if the
sensors were reasonably identical.
A nice to know history bit: already in 1908 some cities turned off the street
lights on clear sky full moon nights to save fuel, because the moon provided
almost as much light as the oil lanterns could.
My concrete thoughts to this tagging idea:
The amount of light and the suitability to that environment (i.e. ruling out
some otherwise bad installations) are the most important assessments for the
quality, and they could be used.
The values should not be loaded with excessive emotions, like "horrible" or
"excellent", but rather we could attempt to define them from what could be
called usability goals of the road users. These are different for different
environments! A motorway user doesn't care about the faces or intentions of
pedestrians, and a pedestrian in a park doesn't need to see if there's a stone
on the road 100 m ahead of them (say, fallen from a truck). For a pedestrian to
be able to see the objects, curb stones and other surface irregularities posing
a trip hazard, the required lighting levels are much lower than what's
comfortable or what vehicular speeds require; rather the actual (statistical)
and perceived safety are considered as key points for a "good" pedestrian area
lighting.
On my travels, I've also noticed that there are big differences in urban
environments' lighting; in places the sidewalks are pretty dark even if the
road is lit, but in others the lighting is more uniform. In urban environments,
the driving lanes are, however, usually brighter. It would be worthwhile to
compare these setups before defining the value guidelines, as to whether the
typical "fit-for-purpose" lighting quality value would be the same for both the
sidewalk and the road, and in which of those different typical environments. It
could be noted here that in some city centers the ambient level is even up to
10 cd/m2, whereas in quieter cities' residential areas the maximum on the road
surface might not exceed 3 cd/m2 and the sidewalk could be even darker; this is
to say that the categorization necessarily depends on the area.
People can, however, visually estimate the amount of light to some degree. At
least the extremes of the scale are easy:
- from soccer field/motorway like lighting with so much light that nothing is
missed because of the darkness
- to unlit footways in a park, near some lit road but with so little light that
you don't see the small surface irregularities. (The next step is where you
can't walk normally because you don't see bigger obstacles, either).
Between the minimum and maximum normally encountered levels, the number of
categories would be difficult to decide, because it's limited by the fact that
people necessarily would categorize different lighting setups to one category
(when they couldn't tell them apart) and the fact that they would sometimes
categorize the same setup to different categories depending on the previous
setup they encountered (i.e. "this is much darker than the previous" from one
direction but "this is almost the same as the previous" from the other
direction).
The first step forward would be a collection of maybe 4 to 7 photos of the
different lighting levels, each with the same camera and aperture/exposure time
settings so that mappers could build reference level mental images. At the
moment my devices don't allow such fixed settings.
I'd leave it up to the mapper to decide whether to split the ways if a small
section is different from the rest of the way. Some guidelines could be
written, anyway; very short darker of lighter sections wouldn't affect the
tagging of the prevalent amount of light. An extra class/value would be
reserved for cases where the lighting seems otherwise good, but the mapper
noticed excessive glare or some other planning mishap that prevents roads users
from assessing the lighting as good as the lighting level would suggest.
In closing, the values should reflect how the lighting works for its purpose:
- for a cycleway (as in "at least cyclists allowed but no cars") the values
could be used to tell if a cyclist can see well enough that they(*) do not need
to slow down because of the quality of the lighting installation.
*) I would exclude "racing cyclists", i.e. count only those "commuting in a
business suit"-speed cyclists
- for pedestrian environments (plazas, sidewalks, other footways
only-for-pedestrians) the objectives of the lighting are to enable safe
movement and to increase personal safety. If it's possible to walk at a normal
pace comparable to daytime conditions, and the lighting is reasonably evenly
distributed and the lighting doesn't mask other pedestrians intentions, it
would be considered as fit for purpose.
- for carriageways with higher speeds, the facial detection is no longer an
issue. I don't yet have a proposal for the criteria of "fit for purpose"
lighting level for these: even the designers don't expect to find the brightest
category illumination on quiet residential roads, but that doesn't make the
brighter ones "excessive", nor does it make the dimmer residential roads
"unfit".
Sections where some aspect of the lighting poses a hindrance, would be marked
as such even if the lighting there was otherwise up to the design standards.
Whatever we would record, it shouldn't be completely subjective ("i like
this"), but it can be something that reflects the objectives of the lighting,
yet is subjectively assessed. The next mapper can correct an assessment, but
they can't correct a feeling. With good examples, the assessments will be
similar enough.
--
alv
_______________________________________________
Tagging mailing list
Tagging@openstreetmap.org
https://lists.openstreetmap.org/listinfo/tagging
