Accession Number:



Using SSM/I Data and Computer Vision to Estimate Tropical Cyclone Intensity

Descriptive Note:

Corporate Author:


Personal Author(s):

Report Date:


Pagination or Media Count:



Satellite imagery and other remote sensing products often provide the only observational data of tropical cyclones. This is especially true in the western Pacific where aircraft reconnaissance missions stopped in 1987. Manual estimate procedures using satellite imagery Dvorak, 1984 provide valuable assistance in determining tropical cyclone intensity. An objective Dvorak technique Velden, et al., 1998 is currently being studied to enhance the manual method. In an effort to take advantage of the unique characteristics Hawkins, et al., 1998 of Special Sensor MicrowaveImager SSMI data, one Naval Research Laboratory effort outside the scope of this paper involves the computation of empirical orthogonal functions of SSMI tropical cyclone data and presenting those values as inputs to a neural network to estimate the tropical cyclone intensity at a given imagery time May, et al., 1997. The algorithm applied in the research described here also uses SSMl data, specifically the 85 GHz H-pol channel and a derived rain rate product. The 512x512 pixel imagery is cyclone-centered and image characteristics computer vision features are computed from the imagery data. A subset of these -features is presented to a pattern recognition algorithm k-nearest neighbor and an intensity estimate is provided as output. A description of the imagery characteristics including available data and computer vision features and feature selection methodology is provided in section two. Section three is a discussion of the algorithm used to automate the tropical cyclone intensity estimate and the current evaluation results.

Subject Categories:

  • Patrol and Reconnaissance Aircraft
  • Meteorology
  • Computer Programming and Software

Distribution Statement: