Automated contrail and cirrus detection using stochastic properties
For the detection of contrails and cirrus clouds, new mathematical methods are developed and applied for different kinds of satellite images (AVHRR, MOS, MOMS, ATSR). As edges and skeletons are not sufficiently representative enough for the description and distinction of clouds, different stochastic properties in their combination are applied. The stochastic properties are represented mathematically by a system of coupled stochastic differential equations. Methods are developed to obtain estimation values from stochastic differential equations. The solutions obtained are given as sequential procedures using grey values of neighbouring pixels, though this is very time consuming so these procedures are approximated for usage of simpler array procedures working on the entire image. For the combination of diverse stochastic and non-stochastic properties, different kinds of properties are represented by generalized measures. This is realized by fuzzy measures and fuzzy functions, relating to selected fuzzy measures. Whereas the fuzzy function describes rather isolated stochastic variations over small isolated regions, the fuzzy measures describe more extended regions by their stochastic properties. Both of these represented stochastic properties that are fused by a fuzzy integral. These results are used as new fuzzy measures or fuzzy functions and generate iterative procedures to select desired properties for the detection of contrails.
No Reference information available - sign in for access.
No Citation information available - sign in for access.
No Supplementary Data.
Document Type: Research Article
Affiliations: German Aerospace Centre, Optical Information Systems, Berlin, Germany
Publication date: 2007-05-01