Digital Geography

unsupervised classification of a Landsat image in R: the whole story or part two

The main question when using remote sensed raster data, as we do, is the question of NaN-treatment. Many R functions are able to use an option like rm.NaN=TRUE to treat these missing values. In our case the kmeans function in R is not capable to use such a parameter. After reading the tif-files and creating of a layer stack we will go on with a work-around to solve the missing values problem of the non-covered areas of a Landsat picture.

unsupervised classification in QGIS: the layer-stack or part one.

The more we work in our special scientific areas and trying to answer often complex questions, we face the problem of the sheer amount of data. Therefore the knowledge about unsupervised classifications, its prior steps and its application in several software packages is the best way to see patterns in your data, extract underlying information or to reduce complexity by using only “relevant” informations of your data. Therefore we use a basic concept: similar things should reflect similar in remote sensed data: an apple and a pear both belongs to fruits and seems to have similar colors in most cases.…