We are often asked questions like, “how do I measure my product using ASD near-infrared (NIR)?” Of course, my standard answer is, “it depends.” Since so much depends on the assumptions and conditions, I thought it would be useful to try to help clarify some of the things that proper material measurement depends upon.
Decide how to collect the spectra
NIR devices begin by collecting a spectrum. ASD’s NIR devices use a portion of the electromagnetic spectrum from 350 to 2500 nanometers. For organic materials, the NIR spectrum is the result of bending and stretching of organic bonds; in the Vis/NIR region these signals overlap. So at any point, the resultant spectrum was influenced by multiple compounds. ASD systems can be used with solar as the illumination source or can utilize accessories with an integrated light source. Collection of a high-quality spectrum is of paramount importance in material characterization with NIR and the extremely high signal-to-noise of ASD systems enables the measurement using various accessories and illumination sources.
In order to properly measure anything, you must decide what it is that you want to measure. Sometimes this is quite easy to understand, such as measuring oil or protein in grains. Other times, what exactly you should measure is not clear. Once you have decided what is important to measure, it is time to find a quality reference method and laboratory to measure these with a reliable and reproducible reference method.
How many samples are needed?
Obtain samples produced by your process or from your sample population that contain the most variation in the constituents that you intend to measure. We generally recommend a minimum of 100 unique samples. A common pitfall is trying to create a model with an insufficient number of samples. After scan collection, samples are sent to the reference laboratory for the determination of composition.
Ready, Set, GO!
Quantitative chemometric models can be developed once you have the spectra and reference data for the samples. Various software packages can be used for model development – most allowing different algorithms to be used. Once the models have been created, they can be used to predict an independent set of samples for validation. After the model has been validated, it can be used with the ASD system to measure hundreds of samples per day to enable control of your process or characterization of your materials. Better characterization means more efficient operations.
The good example
A good example of this process was published by a group of researchers in China. “Non-Destructive Evaluation of the Leaf Nitrogen Concentration by In-Field Visible/Near-Infrared Spectroscopy in Pear Orchards”
This paper evaluated leaf nitrogen content by reference methodology and scanned leaves using an ASD FieldSpec system with a 25 degree foreoptic and using ASD plant probe and leaf clip accessories. Fertilization of important orchard crops impacts the yield of these crops and by using an ASD Fieldspec system to evaluate leaf nitrogen, they were better able to establish orchard management practices to optimize nitrogen application. They collected data over a two year period and created models using solar illumination and two different backgrounds with the ASD plant probe and leaf clip. Multiple calibration development algorithms were also evaluated in this paper.
I encourage everyone to read this excellent paper and my congratulations to the authors on this outstanding use of an ASD spectrometer system and a very well planned design of the experiment!