Cloud Detection

To improve the definition of useful areas under clouds, it is necessary to detect cloud and shadow in images.

This tool is intended to speed up the cloud and shadow detection process by means of image processing specific techniques applied together and following pre-defined pattern recognition resulting in a vectorial data representing analyzed objects.

Different from other algorithms for cloud/shadow detection, this tool is able to detect these objects in an image of any satellite, not being specific to some sensor type. It is also possible to use some image extension types, such as TIFF, JPEG, etc.

To detect such objects it is necessary to define which image bands will be analyzed, i.e., the analysis is done in distinct bands of the image because each image band has a better response to certain targets. For instance, LANDSAT satellite images have a better definition of high radiance objects such as clouds in the blue band while in near infrared band it has a better definition for low radiance objects such as shadows. 

It is acessible throught:

Processing-> Raster Processing-> Cloud Detection...

Input Layers Histogram and Minimum/Maximum values

Morphologic Operator

Morphologic operators (such as Dilation and Erosion) can be applied to clean the output mask generated by cloud/shadow detection. Small objects such as spikes or holes can be removed by applying Closing/Opening filters.

Output

Ref: http://marte2.sid.inpe.br/col/dpi.inpe.br/marte2/2013/05.29.00.32.13/doc/p1188.pdf