Automatic Contrail Detection and Segmentation in Polar-orbiter Satellite Images

Open Access
Padmanabhan, Viveknarayanan
Graduate Program:
Electrical Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
March 21, 2014
Committee Members:
  • Andrew Mark Carleton, Thesis Advisor
  • Timothy Joseph Kane, Thesis Advisor
  • William Evan Higgins, Thesis Advisor
  • Contrail
  • Line detection
  • segmentation
  • Hough Transform
Contrails are important in local to regional scale climate change. Various studies to date have found circumstantial evidence linking contrails with modifications in surface temperature (e.g., in the eastern U.S. and parts of Europe). It is clearly important to understand what a contrail is and where and when contrails occur to more definitively relate their occurrence to surface climate. The use of surface observations for developing contrail climatology is problematic, owing to the occurrence of intervening cloud layers. Accordingly, satellite images have been used to identify and map contrails, mostly from manual (subjective) interpretation of images, although some automated (quantitative) approaches have been developed but whose success is variable and dependent on a number of factors (Cirrus clouds, curved contrails, complexity of algorithm, etc.) Detecting contrails, therefore, is critical in understanding the atmospheric effects of aviation. This research involves the automatic detection of jet contrails in Advanced Very High Resolution Radiometer (AVHRR) imagery with a high degree of confidence and its segmentation written in MATLAB programming language. Contrails are characterized as thin, nearly straight linear features of higher intensity than the background. Contrails possess another highly characteristic feature; they tend to create straight lines in satellite images. Due to the large volume of satellite imagery, selecting contrail images for study by hand is impractical and highly subject to human error. It is far better to have a system in place that will automatically evaluate an image to determine whether it contains contrails and where. This research develops and tests two new and easier quantitative approaches to find contrails in satellite image data, for a variety of atmospheric and cloud conditions (e.g., clear-skies, partly cloudy skies; cloudy skies).