[Radiance-general] Radiance, analysis

[Rob Guglielmetti]

steve michel wrote:
> The day time illuminance pic shows the lens in black in the obviously much brighter pic (I adjusted exposure) but the ximage 'l' readings are too high for fc units but seem correct if they are lux. The key word is 'seem'. 
>
> My variables are too numerous to list here( rpict options, ximage etc.) I need to know if an irradiance pic rendered with thr -i option in rpict can be analyzed ximage so has to give lux or footcandle readings. There is sun patch in the daytime rendered scene. would that serverly impair the accuracy of pixel readings???
>   
Hi Steve,

By default, your luminance pics are in cd/m^2, and your illuminance pics 
are in lux.  The accuracy of the rendering would not necessarily affect 
the accuracy of the scene; if anything, sun patches would improve the 
accuracy of the results for a given set of settings.  I wouldn't worry 
about the appearance of an illuminance image, if the values are 
correct.  I recommend you use the falsecolor tool to generate some 
falsecolor images of your scene, so you can do a thorough investigation 
of the whole image's values.  And of course for absolute accuracy of a 
validation you should use rtrace and calculate values at specific points 
that you can verify with hand calcs.  This will be more accurate than 
interpolating a glob of colors in a falsecolor image, or even picking 
points off the image in ximage. 

Also, have you tried rad yet?  rad can make very intelligent guesses 
about appropriate settings for rpict, based on your simple inputs of 
scene size (max dimension), image QUALITY (high-medium-low), VARIABILITY 
or range of light in the space (high-medium-low), and DETAIL or model 
complexity (high-medium-low).  This is a great tool and you can even use 
it and just output the settings to a text file to inspect.  You can 
learn about the variables that way, by changing one of the parameters 
above and see the ramifications on the rpict/rtrace settings.  The use 
of rad is covered well in RwR, as is the practice of performing 
sensitivity studies on your models and parameters, to find a sweet spot 
in terms of accuracy/time.  This dance is done on every project, but 
eventually you will develop a feel for setting these parameters, at 
least roughly. 

- Rob Guglielmetti
www.rumblestrip.org