Preprocessing Ground-Based Visible/Near Infrared Imaging Spectroscopy Data Affected by Smile Effects

Henning Buddenbaum, Michael S. Watt, Rebecca C. Scholten, Joachim Hill

Research output: Contribution to JournalArticleAcademicpeer-review


A data set of very high-resolution visible/near infrared hyperspectral images of young Pinus contorta trees was recorded to study the effects of herbicides on this invasive species. The camera was fixed on a frame while the potted trees were moved underneath on a conveyor belt. To account for changing illumination conditions, a white reference bar was included at the edge of each image line. Conventional preprocessing of the images, i.e., dividing measured values by values from the white reference bar in the same image line, failed and resulted in bad quality spectra with oscillation patterns that are most likely due to wavelength shifts across the sensor's field of view (smile effect). An additional hyperspectral data set of a Spectralon white reference panel could be used to characterize and correct the oscillations introduced by the division, resulting in a high quality spectra that document the effects of herbicides on the reflectance characteristics of coniferous trees. While the spectra of untreated trees remained constant over time, there were clear temporal changes in the spectra of trees treated with both herbicides. One herbicide worked within days, the other one within weeks. Ground-based imaging spectroscopy with meaningful preprocessing proved to be an appropriate tool for monitoring the effects of herbicides on potted plants.

Original languageEnglish
Article number1543
Pages (from-to)1-11
Number of pages11
JournalSensors (Basel, Switzerland)
Issue number7
Publication statusPublished - 30 Mar 2019


  • Specim FX10
  • field imaging spectroscopy
  • forestry
  • herbicide
  • invasive species
  • preprocessing


Dive into the research topics of 'Preprocessing Ground-Based Visible/Near Infrared Imaging Spectroscopy Data Affected by Smile Effects'. Together they form a unique fingerprint.

Cite this