SEA ICE SURFACE / MELT POND INVESTIGATIONS
BY MEANS OF HELICOPTER BASED DIGITAL CAMERA RECORDS
This example is dealing with investigations of melt pond properties that were derived from Helicopter based camera records during the AWI's ARK17 expedition in summer 2001. Before the flights at 5 different days, a 3 channel (RGB) digital camera was mounted below the Helicopter in order to collect videos of the sea ice surface while the platform is moving with constant speed at constant height. Extracting images from these movies in a way that the alignment of these images represented the according flight track, we obtained around 10 000 images of Arctic sea ice that we could use to classify melt ponds and analyse their properties.
[SASCHA WILLMES, Mar/Apr 2002]
Example
(left column: photograph; right column: binary melt pond classification)
Flight No.237, height: 90 m, flight speed: 40 m/s, only melt ponds highlighted
PROCEDURE
It is easy to see that it is quit easy to distinguish open water from sea ice by using simple gray value tresholds. But still, there is a lot of unclassified transition gray values that can not just be included to the melt pond class since it also contains shadows of pressure ridges.
So another criterium has to be found to get rid of the dark shadow pixel in this gray value transition so that we can just place melt pond class between the ice and open water treshold.
Looking at the average reflection of all components from the blue to the red channel, we see that shadows and melt ponds nearly cover the same gray value region but differ in relative changes from blue to green. This was used to isolate shadow pixel by just subtracting the green from the blue image, enhancing the contrast and binarizing. Not all shadows could be found by this method but anyway the result was really good for this uncomlpicated method.
