Hello,
I have spectral reflectance data coming from a plant. As the figure shows, there is noise present in the data around the 1500 and at the end. We do not know a priori where this noise will occur because it may depend on how humans manipulate the corresponding device.
Is there any standard algorithm (hopefully with R code) for removing this noise automatically?
It really looks like absorption bands from water: one 'small' band around 1500nm and another strong one starting around 2000nm. (look at the bottom website for example).
Just an observation...
Here is the link...
http://www1.lsbu.ac.uk/water/water_vibrational_spectrum.html
Could try the prospectr package that implements a number of algorithms and techniques, not just signal processing but also calibration sampling, related documentation here:
https://cran.r-project.org/web/packages/prospectr/vignettes/prospectr-intro.pdf
For computationally intensive work involving larger data set, there is an interesting approach proposed by Guangwei, Haotong & Zhongrui (2015), involving the extraction of 1D spectra from 2D spectral images: http://arxiv.org/abs/1506.06188
I don't know how the measurement is made, but if it involves the division of a sample measurement vs a reference measurement, it is not impossible that absorption due to humidity in air will give rise to this kind of noise. (Without having to take into account the influence of the water content in the plant itself.) My 2 cents.
If you have a series of measurements, may be calculating confidence interval for every bunch of curves ? For instance, I had a series of data of water spectral reflectance (also contained noise), then took into account those values within the confidence interval or mean values... But I am not quite sure about your case: trees canopy is of course another case. I worked with seagrass and water column spectral reflectance a while ago and processed data in Gretle/GNU. Of course your case is different. But just if of any help:
Thesis Seagrass Mapping and Monitoring Along the Coasts of Crete, Greece
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs