Landsat images have 30m spatial resolution as we all know. Which is why I have faced difficulty in differentiating different Land Covers such as: forest, vegetation, agricultural land as the resolution of Landsat Images are not fine enough.
I am asking if there is any method to detect areas in more specific scale? (except panchromatic correction)
Dear Gazi
Here is simple and good research paper attached where authors used landsat satellite and categoried Agriland,, Forest, mix vegetation etc. You may follow the proposed methodology
Try this article.
http://digitalcommons.unl.edu/cgi/viewcontent.cgi?article=1538&context=usgsstaffpub
you have a powerful free images from sentinel 2A with a large number of Near infrared channels, you can easily differenciate between vegetation.
You can usually distinguish the three cover categories directly by visual interpretation depending on cropping season, deciduous versus evergreen forest, coniferous versus deciduous forest and crop field size. An unsupervised spectral classification will also usually produce the three categories. A supervised classification will be better, but obviously needs some ground or other secondary data on the targeted cover types in your AOI.
You can use the spectral signature curves to identify the different types of veg. ENVI library has spec from different veg types, you just compare the ones from your image to the once in ENVI library. You can access the ENVI library by clicking on Spectral/ Spectral libraries, in the dialog box that appears, click on the "Open" tab below the box and click on spectral libraries. Normally these files will be under the C:/program files and either Exelis or ITT depending on the version of your ENVI software, then ENV50/ resources/speclib. (Alternatively just go to C drive and under Envi software search for speclib.
I hope that helps
Although all the material you were already given looks pretty nice, what you need to do first is to investigate the local geology, crops and flora. Once you have an idea of the kind of crops and rocks, and independently of the software you are using for the classification of your multispectral images, you need to study the spectral signatures of your target materials, for which I usually use this webpage: http://speclab.cr.usgs.gov/spectral.lib06/
Then you need to analyze the peaks of your targets in terms of the bands of the satellite images that you are using, since the sampling resolution of the signatures will differ. You will realize that even though the signatures of, let's say, grass and pines are very similar, they have slight differences in the absorbance peaks. You can use those differences to help your classification, for which you can use a "supervised classification", in which you need to create "training polygons" defining the areas where your target materials are. You should be very careful during this process, so that you isolate very well the target materials. I use polygons of 10-30 pixels, and as distributed in your study area as possible. I need to mention that a good supervised classification should also include training polygons for the materials that are not your target, i.e., the backgroud of you study area, such as bare soil, water bodies, etc. What you are doing requires a strong understanding of all the materials in the area, and I can tell you by experience that Landsat can be used to perform such classification, with the limitations of the spatial resolution of your pixels.
I hope this short explanation can help you.
Dear Gazi
Here is simple and good research paper attached where authors used landsat satellite and categoried Agriland,, Forest, mix vegetation etc. You may follow the proposed methodology
- Badges
- Science topic
- Similar topics
- Geoscience
- Geoinformatics