After some days of hesitation, I have finally managed to complete my first land cover classification.
To be honest, I had already reach a result months ago, but the result wasn’t what I expected.
First of all, what is a land cover classification ?
A land cover classification is a process in which we use a satellite frame (a raster), that we convert into a land cover image, and then a lot of land cover polygons (vectors), with which we can work.
When you are downloading your first satellite images bundle, you will discover in your file, an significant number of files, and images. Those images are the different bands, that we will combine, to create a composite image, with which we will be able to work. Each composite image will help us to better see the differences between a forest, a city, a river, an irrigated crop land, and then well classified everything (because even if our software can help us, we will first train the software by hand!).
Training my software had been my major issue to be honest. To train it, we have to circle, draw polygons on our composite image, telling the software that on these pixels represent either a forest or a building. Obviously, you won’t have to do it pixel per pixel, but you will to try to be the more precise possible.
After the completion of that task, the software will process the datas and…. Magic, you will have a first sight of the classification. After cleaning your file, you will be able to convert into a vector file, and at this moment, treat the data you want.
That is not an easy task, at the beginning, but with training and courage, you will do something nice, that as precise and correct as possible !
