Landsat 8 Image Classification with ArcGIS (Supervised)

hi welcome to another tutorial in this tutorial i'm going to show you how to do a supervised classification in order to turn a satellite image that looks like this into something that looks like this this classification actually has been done to a selected area of your interest based on the downloaded landsat 8 images using arcgis supervised classification module as you can see over here these red color regions actually referring to urban areas and the blue color ones are obviously water bodies and the light green corresponds to agricultural areas the yellow patches are actually bare lands and also the dark green areas correspond to forested areas so without further ado let's go ahead and get started with the tutorial so the data that i'm going to require for this tutorial is basically going to be a landsat image which i'm going to download from this earth explorer web portal you can log into it using earthexplorer.usgs.gov and as you can see over here i have my my area of interest is actually located somewhere in the middle of france in the central region of france based on my criteria as you can see over here i have selected one satellite landsat image and during the process of setting the criterias you can actually set the criteris to have an image with a less cloud cover and if you can get one of those i think we will have less troubles during our classification process so once you have selected the corresponding image which you would like you can just go here and download but just make sure that you actually create one account with the with usgs first because if you don't do that it won't let you download the image all right so after you download the image you can actually open arcgis and then you can navigate to the place where you have saved your landsat 8 image the downloaded one and in my case you can see that inside this supervised classification i have a number of files which actually gets downloaded when i download the whole thing you can see that we have bands corresponding to band number one two three four five six seven eight nine ten eleven and so on and if you're a little bit confused right now what each of these bands uh refer to uh you can have a look at this table and as you can see each of the bands are actually corresponding to different uh different wavelengths now we have the band number one which is the coastal aerosol band and band number two three four are blue green red and band number five is the near infrared and so on and and for this classification process actually we are going to only use bands from one to seven so even though you have the whole set of bands over here you might not need to actually select everything so just go ahead and select only from band number one two three four five six seven all right now after a while you'll be able to see that we successfully imported all the bands that we would require for this exercise band number one two three four five six seven all right now what we are going to do is we are going to create one composite image what that basically means is that it's a it's a combination of all these bands together so in order to do that you can first go to windows and go to image analysis and then you will see that this image analysis window gets opened from here you can just resize this and then just make sure that you are selecting all the bands all the way from one to seven and after you do that just go to this composite bands button and once you click that you will see that there will be a composite band which gets formed on top over here all right now this is the composite image that we're actually going to work with so basically we won't be requiring any of these individual bands anymore so what you can do is just to make your workspace clean you can just go ahead and clear this out by selecting everything and then you can just remove them all right now this is the composite image that we created now since we created this composite image using seven different bands we actually have the freedom to sort of uh select different combinations of different landsat bands now as you can see this table shows you the different band combinations for landsat 8 now the good thing with using this multi-band landsat images for image classification is that you don't have to actually find out different land use classes just by looking at the image you can actually get different combinations of bands to sort of guide you and give you give you some sort of an indication just to make it easy for you to identify what each landis class might refer to the things which actually are displayed in image what do they mean in reality for example if you would like to identify uh maybe the water bodies or if you would like to identify or distinguish between the forest areas and the agricultural areas it might be a bit hard if you're if you're trying to do it just by looking at the image just by looking at a natural color image so that's where this kind of a table actually this kind of band combinations actually come in handy now for example let's get started with this band combination 432 now 432 actually refers to the natural color composite so what you can do is just go over here to the red band and click over here and you can select four and the green band select it to be three and the blue band selected to be two now this is called actually a natural color image because the colors that we can see from this band combination is actually quite equivalent to what we would see just outside you can see that we have some green areas and then also if you're looking at let's say if you're looking at a google earth image you will see that your water bodies might be like a bit dark over here you have something which which appears to be something like an urban area over here and these brown patches probably might be referring empty lands or something like bear lands so these are the colors that actually we are used to seeing so that's why this uh 432 band combination is is known as the natural color band combination now let's see if you would like to have a very sharp distinguish between the land areas and the water bodies a very well known band combination for that purpose is actually 564 so you can assign five over here six over here and four over here and when you do that you can see that the water bodies are actually getting appeared in a very dark dark sort of a blue color and the bear land you can see that the bear land is actually getting displayed in sort of yellowish color you can see the dark yellow colors actually referring to uh land areas where there's nothing just maybe nothing but just just plain soil so in this way you can very easily distinguish using this sort of a band combination now another famous band combination is actually 543 which is also known as the color infrared band combination which can be used to identify vegetation now let's go ahead and do five four and three over here now when you set this band combination the more dark patches that you will see actually are referring to very sort of dense or healthy vegetation now you can see that these areas probably are corresponding to forested areas that's why it appears in like very sharp very dark red and there are certain other areas that actually appear in sort of a light light color red which might probably correspond to agricultural areas so in this way you can actually further confirm what you might actually be looking at using the satellite image now similarly if you would like to identify the urban areas now you can see that the urban areas sort of appear in sort of a very uh whitish blue over here even these areas i think can be attributed as urban areas so i guess you get the idea how you can use how you can use these different types of band combinations to your advantage in order to identify certain land use classes all right so i'm just going to revert this back to the natural color image which is uh 4 3 and 2. all right i'm going to get started from here but in the middle of the classification process i might switch back to different band band combinations so before moving on to the classification step let me just go ahead and save this image you can just save this composite image by right clicking on it go to data and export data and from here you can navigate to the place where you would like to save and you can also give the file name and just make sure that you save it as a tiff file maybe i'm i'm going to name this as composite image yeah as you can see it might take a while because this is a bit of a heavy file okay now the file has already been exported so i'm just going to go ahead and get rid of this first yeah and to begin the classification let me put this back to the natural colors that means band four three and two yeah this is the combination now you can see actually this is one this is one tile of a set of a landsat satellite image now i have an area of interest which i actually would like to focus on so i'm going to import that shapefile which specifies the region of interest so for the classification i'm first going to actually do the classification for the whole tile and later when i create the rust i'm just going to perform uh extract by mask operation just to actually extract only the area which i need but for the timing i'm just going to mainly focus on the on the areas which falls within my bounding box so now what i'm going to do is i'm going to change this uh this bands from 432 to 564 when you use this band combination it actually lets you very easily distinguish between the land area and the water bodies now as you can see the water bodies actually gets highlighted gets colored by by a very dark blue almost it looks like black but this very dark blue areas actually corresponding to the la to the water bodies so now what i'm going to do is i'm going to actually provide samples now in order to provide samples what you can do is you can right click somewhere over here and make sure you activate this image classification option right once you have that you can open this training sample manager and now you can go to this option and you can specify the samples either by drawing a circle or you can draw a rectangle or you can even draw a polygon so in my case i actually choose to draw a polygon all right so now what you can do is you can specif you can zoom into the areas zoom in very well just to capture as many samples as possible like this which are corresponding to water bodies because that's the first class that i'm going to provide the classification for so maybe we can provide two or three samples from the river i think it's not going to make much of a difference all right and apart from that i'm also going to try to find a water bodies which are located within the land such as these areas and this looks like a water body to me as well and maybe this one all right now i have provided six samples now what i can do is i can actually group everything together using this first you have to select all the classes that you all the samples that you provided and then you can go ahead and click merge training samples over here and then we can give a name for this class motor bodies maybe you can change the color as well to be blue all right so the next class that i'm going to classify is actually the bear lands now if you can recall i just mentioned that you can even use this 564 band combination to identify to distinguish between the water bodies and the bear lines because the water bodies will appear in a dark blue color while the bear lines actually will appear in sort of a yellowish orange color also if you would like to actually look at the natural color image and do the classification that's also possible you can just turn this back to 432 and again from here you can actually clearly see where the bear lines are so i'm just going to do the bell and classification based on maybe based on the natural color image i think i can probably select a sample from here as well as you can see that it looks like some sort of an open ground or maybe some some kind of a terrain which actually doesn't have any vegetation on on top of that maybe i'll still actually call that as bear land as well we'll give maybe another sample from here all right so now i'm going to group all of these together and specify that as bear land all right now my next objective is actually to identify the urban areas let me go ahead and give this the sort of an orange color now to identify the urban areas i would say that i would rather actually switch back to the five five four three combination if you can recall it five four and three this is the color infrared band combination which is also called as the near infrared composite now in this one the urban areas actually appear to be in sort of a bluish white now as you can see over here all of these are actually urban areas so i'm just going to classify these points these areas as urban areas so i'm just going to provide the samples first yeah i guess that should be that should be okay now i'm just going to go ahead and select everything together and this one i'm going to name it as urban areas and i can still use the same band combination actually to identify the vegetation now as you can see over here there are two types of reds over here you can see that there's a dark kind of red and then there's also a sort of a shaded shaded red now these dark reds actually referring to healthier dense vegetation which probably could mean something like forested area or something like that so i can again still take some samples from here in order to specify the forested areas and maybe a couple of samples from here as well and this one i'm going to group and specify that to be forest areas which should be represented by a darker green color i guess and now i'm going to switch back to the band combination 6 5 2. which actually also let's sort of uh double confirms our previous selection about the forest areas now you can see actually there's a clear distinction between this dark green and the and the light green over here so this dark green actually refers to the forested areas quite possibly and these light green areas i'm going to name those as agricultural areas so let me go ahead and take a couple of samples over here yeah i think that should be sufficient let me go ahead and group everything together all right agricultural areas and as you can see i have now five land use classes so i think this should be sufficient for the purpose of this tutorial now if you actually want to do a much more detailed classification you can always play around but i guess you you get the main idea of what i'm trying to do over here now yeah now before moving on to the classification in case if you need to change something in this training sample manager what you can do is you can simply select that and select that item and then you can simply go ahead and either delete it or you can actually expand it from here so for example if i want to reselect the urban areas what i can do is i can just delete this training sample and then i can start selecting again now let's say if i want to select the urban areas i would like to maybe flip this back to 543 and from here maybe this time i'm going to select the areas which appears to be something like rooftops areas with only a white color signature let me try that all right now you can again merge everything together and maybe assign urban urban areas and this time maybe i'm going to select it in red color i guess all right so once everything is done once you are sort of confident about your classification which you can always come back and change you can just go to this classification and select interactive supervised classification option all right now you can see that based on my provided samples this is how the classification actually looks you can see the urban areas are sort of confined to this area and then you can see the water bodies have been classified but in some places you can see the classification actually has not happened properly now for example over here you can see that the river is actually a bit discontinuous now this is due to the less number of samples that we have provided so in if you if you in case if you come across an issue like this what you can do is you can actually go back to the water bodies and you can add one more sample or you can if you would like to actually sort of re-sample the whole thing again you can just get rid of this and then a higher number of samples which will actually provide the program with the higher number of spectral signatures sorted associated with different types of water bodies so and if i go to this western side you can see that actually the forested areas are located mostly in this region as you can see from the image as well all right now what i can do is i can actually extract the area which i need using the extraction by my extract by mask tool so that i can actually get rid of all the surrounding areas which i do not really need so for that you can go over here go to this search panel and search for extract by mask and from here you can drag this and drop it over here let me close this one now and from here i'm going to select my region of interest all right now i can close this layer out all right now you can see that we actually have extracted the area and then have classified it accordingly which shows us the water bodies the bare land the forested areas the agricultural areas and also the urban areas which are confined actually to this uh to this region now this would be actually a good stop to conclude this tutorial but i'm going to go one one step ahead in case if you would like to calculate these areas let's say you would like to know inside your region of interest what is these what is the amount of square kilometers of urban areas that you have so one way you can do this is by converting this into a vector so because this is right now this is still a raster so what i can do is i can convert this raster into a polygon convert this raster into polygons using raster to polygon tool and i can select this and now if i open the attributes table over here you can see that i can assign the classes names name on the class name that i have provided over here so in the field i'm going to select last name and also i'm going to simplify the polygons and now i can just go ahead and press ok now you can see actually we have the polygons corresponding to each land use class now now as you can see over here it's actually it has been split into each small pieces now i'm going to sort of dissolve everything based on the class name so what happens in that case is that let's say if you want to group everything that belongs to the agricultural lands it actually groups everything together so what you can do is you can just go to geoprocessing options and go to dissolve and you can select this and dissolve it based on based on this class name so i'm going to select the class name over here and this time i'm going to specify the path let's say land use classification all right now i'm going to go ahead and click ok all right now if i open the attributes table you can see that it has been grouped according to the class name now if i select over here you can see that all the agricultural lands will get selected and similarly the bear lands will get selected the forested areas the urban areas and the water bodies alright now what you can do is maybe let me go ahead and get rid of these other unwanted layers now again if you would like to assign proper color proper classific classified colors for this uh each land use class maybe you can go to properties and you can go to symbology and go to categories and now we are going to categorize it based on based on the class name so i'm just going to say add all values and maybe from here you can manually change the way how you would like it to appear let's say the agriculture is a dark green bell and maybe i'm going to retain the yellow color the forested areas sorry let me go ahead and change the agriculture to be a light green and forested areas may be a dark green urban i can retain red and water bodies of course we would like to have blue yeah now you can see that it actually got the colors got assigned properly you can even go to the properties of all symbols and you can get rid of the outline color over here so that it'll actually appear in the way that you would like to have all right now this is not a raster this is a polygon so you can actually do all the operations that you would normally do with polygons such as finding the area so what you can do is you can add the field and maybe add a add an attribute call area and maybe set it to be double and from here you can right click and go to calculate geometry and from here you can select the unit i would like to have the areas in square kilometers and now when you click yes you can see actually what are the areas corresponding to each land use type maybe i can go ahead to the properties and go to numeric maybe reduce the decimal points to let me have no decimal points i guess yeah yeah from here you can see that from the whole area it has 1084 square kilometers of for of agricultural areas 324 square kilometers of bear land forested areas urban and only about 10 square kilometers of water bodies so i guess this concludes the tutorial if you do have any questions you can definitely comment them down below i guess you'll learn something new and if you did like the tutorial kindly show your support by hitting the like button and also consider subscribing if you would like to get more cool tutorials like this uh in the coming days so thanks a lot for watching guys i'll see you in the next one

Channel: GeoDelta Labs

Views: 80,947

Rating: 4.9603729 out of 5

Keywords: arcgis, arcmap, landsat 8, supervised classification, qgis

Id: x8uFTix3sHU

Length: 27min 53sec (1673 seconds)

Published: Fri May 01 2020