Artificial intelligence controller for automatic multispectral camera parameter adjustment

Haitao Xiang, Lei Tian

Research output: Contribution to journalArticlepeer-review


For machine vision and remote sensing systems, image quality is the major factor directly affecting the final image processing results. Especially under natural lighting conditions, image color varies significantly as the lighting conditions change. Unfortunately, changing lighting conditions are inevitable in crop field remote sensing applications. This study was undertaken to develop an artificially intelligent controller based on an artificial neural network (ANN) and an adaptive neuro-fuzzy inference system (ANFIS). The aim of the implemented controller is to automatically adjust multispectral camera parameters, such as "gain" and "exposure time," to compensate for changing natural lighting conditions and to acquire white-balanced images. A high-resolution digital multispectral camera was used as the image sensor. A calibration panel with 56% reflectivity over the visible and near-infrared bands was used to provide the white-balance reference. The real-time white-balancing of the image was achieved for three channels of the camera under natural lighting conditions. It was shown through experiments that the developed algorithm was able to complete multispectral camera parameter control within three iterations for each channel. The convergence speed was faster than with conventional control methods.

Original languageEnglish (US)
Pages (from-to)1873-1881
Number of pages9
JournalTransactions of the ASABE
Issue number5
StatePublished - Sep 2007


  • Adaptive neuro-fuzzy inference system
  • Artificial neural network
  • Machine vision
  • White balance

ASJC Scopus subject areas

  • Forestry
  • Food Science
  • Biomedical Engineering
  • Agronomy and Crop Science
  • Soil Science


Dive into the research topics of 'Artificial intelligence controller for automatic multispectral camera parameter adjustment'. Together they form a unique fingerprint.

Cite this