Hi
I'm wondering if anyone has own experience or knows a good-quality experimental work, where the topic is explored? I have some repeated airborne +/- 30 degree scan angle (at 1550 nm) waveform data, which I need to calibrate (or rather verify that the sensor behaved similarly in the repetitions). I'm observing that while a gravel surface shows almost no effect (on top of spherical losses), signals from old dry and bright asphalt decrease drastically at scan zenith angle > 20 degrees.
I have seen some results indicating that SZA effects (in LiDAR) indeed vary a lot, but haven't seen a case with asphalt among the tested surfaces, and verification of the sensors linear response (intensity is a ratio-scale measurement of instantaneous at-sensor radiance) that is needed to be able to eliminate the component by spherical losses. This is what I mean by my notion on good-quality work.
BR, ilkka korpela, U Helsinki
You may try to use statistical methods by using bins of distance, angle and roughness for all points related to asphalt. Here is a paper that was used to identify layers in outcrops from LiDAR scans, but the same method may help or could be used to identify asphalt as a function of an angle of incidence. Here is the link:
"Using Statistical Methods to Correct Lidar Intensities from Geological Outcrops"
http://www.hitechmex.org/publications/asprs2009.html
Hope this helps you in your research.
Hi Reuben and thank you. I'm trying the physically based approach (with radar equation), although not all variables (influence of the medium, 700-1000 m of atmosphere at 1550 nm, which you point out) are under control. Data-driven (yours) is of course an option. I'm not trying to detect asphalt, but using it as a reference to analyze if three repeated airborne scans (spanning 38 months) match (that really the instrument was unchanged, its receiver and transmitter). My scene is 95% forest (trees being my target of interest) and there are very few flat-well-defined-hard targets to serve the purpose of calibration. Can't use the trees for that as I'm studying changes in them. What I have note is that scan zenith angle (incidence angle) effects vary between say for example gravel and asphalt roads. I'm almost sure that my multitemporal "intensity" data are on the ratio scale (measurements of instantaneous at-sensor backscatter radiance) so that I can use the radar equation for the compensation of spherical losses without the danger of causing an artifact. So, the study that I'm looking for needs to have it proved that the "intensity" data are on the ratio scale - that's when the incidence angle effect observations (statistical effects) are free from compound effects. I have waveform data, and do not use the term intensity as its sometimes ambiguous.
In your paper you start with the notion of instruments correcting the degradation by distance - that I call spherical losses - and the issue with forest vegetation is that its ill-posed - because of the geometry of scatterers. If "intensity" is a measurement of return energy (integral) instead of some (instantaneous) peak value observations - this distance (or range) correction will work even in vegetation (soft targets) if a quadratic correction is applied.
I just wish to know if I observe similar incidence angle effects in asphalt as others have. Then again, there are some many different types of asphalt that this may be not worth the trouble.
ilkka to Reuben Reyes
- Badges
- Science topic