How we build spectral models, what results from specific applications show, and what’s new in VIS-NIR spectroscopy. Articles, measurement data, and scientific publications from the ScanSpectrum team.
From choosing reference samples to cross-validation. A practical step order for VIS-NIR calibration, and the most common reasons a model stops working in the field.
Four transforms that decide a model's quality more than the choice of algorithm. When each one helps, and when it strips out the information needed for prediction.
Prediction accuracy for nitrogen content across four growth stages, compared against lab analysis. Where the model is confident, and where it needs extra calibration.
An admixture of cheap rapeseed oil changes the spectrum in a few narrow ranges. We show at what admixture share the measurement picks it up.
A summary of the ScanSpectrum team's paper, with a link to the full text and dataset. Four seasons of field trials, 18 locations.
Between 690 and 750 nm, leaf reflectance rises several-fold. This one edge carries more information about plant condition than the rest of the visible spectrum combined.
Two hundred years from splitting light to a device that measures a material's composition in the field. The milestones that led to the method's availability today.
A new entry is one Markdown file in src/content/blog/<locale>/. Frontmatter describes the title, category, kind, date, read time, and lead; the rest of the file is the article body in Markdown — ## headings are picked up automatically into the table of contents.
The entry fields match the collection schema defined in src/content.config.ts: title, category, kind, date, readTime, description. The same set is enough to configure the collection in PagesCMS, so editors can work from a panel instead of editing files directly.