Findclouds Manual 32
in cloudiness.
When thinking about these problems, it took our attention, that there are characteristic patterns of
sunpixel, which change with amount of cloudiness or atmospheric opacity. Especially sunpixel of
cloud-covered sun often contain gaps resulting out of covered parts of sun. So a simple pattern
recognition was implemented to improve evaluation of sun.
"Sun Objects", "Cloud Objects":
The algorithm analyses the sunpixel. It determines connected
areas of sunpixel (called "Sun Objects") and of skypixel that are included by sunpixel (called
"Cloud Objects"). Than it counts the number of pixel they consist of.
More than one Sun Objects is evidence of a cloud covered sun. Because of sparks, the number of
Sun Objects often is not sufficient to make a reliable statement.
( Sun Object >1 → sun cloud covered )
( Cloud Object >2 → sun cloud covered )
Example:
20130928120000_0
This is an example of an image, which is difficult to analyse.
At first glance there seems to be a visible sun. When taking a closer look, a small cloud in front of
the sun is detected. This highly enlightened cloud gets part of the sunpixel and changes the
pattern like an unclouded sun would produce. Also the cloudiness results in a large area of
brightness.
Sun Objects: 11749 98 28 9 9 8 6 6 5 5 5 3 3 3 2 2 2 2 1 1 1 1 1
The algorithm finds 23 Sun Objects, one of them containing most of the sunpixel
and some "sparks".
Cloud Objects: 50 19 11 5 1
The algorithm finds 5 Cloud Objects. At least two of them contain a considerable number of
skypixel. This is strong evidence of a cloud covered sun.
"Sun Object Ratio"
: Is a characteristic value to determine the state of the sun. It is determined by
dividing the pixel number of largest Sun Object through pixel number of all Sun Objects. Normally
a Sun Object Ratio larger than 0.9 points to an uncovered sun. ( >0.9 → SunOnClearSky )
