QGIS User0014 - Multi Spectral Imagery

hi in this video I'm going to take a look at multispectral imagery so multispectral imagery is mostly satellite imagery but there are even UAVs or domes that can have a multispectral camera cameras and they are really useful for agriculture for instance and we are going to look at one example of that a bit later but first of all we are going to take a look at what what are this multispectral imagery and how can we use it in qgs there are more advanced plugins and that are specialized on using multispectral imagery but they do have a pretty steep learning curve but if you are a professional or want to use multispectral imagery professionally i really suggest you take a look at those plugins and I will show you one example in just a little bit so first of all let's see if I have the right document selected multispectral well what it is is that you don't just limit yourself to the red green and blue color channels or light wavelengths so you have rosters with bands for separate light wavelengths and there are as I said not just in the red green and blue spectrum so you can for instance have ultraviolet infrared and a multitude of frequencies or wave lengths and you can combine these in a way that gives you our true color image and we will do that in just a little bit but you can also use them to calculate more scientifically for different types of results so let's see I'm going to use Copernicus data from The Sentinel satellites but you could use Landsat imagery or any multi spectral imagery that you have access to the Copernicus portal and here is free to use but you need to register for an account to download but there are multiple sources online I'm sure you can find a suitable one and that suits your needs so for this I have just selected an area and southern France for no reason at all other than was a bit curious and it gave me access to images that are reasonably cloud free and current so I'm recording this on the 24th of May and the images I'm going to use is taken yesterday so they are really current images but as you see when I scroll down here in the available imagery for this selected area some are really cloudy and that is the flipside of the coin when you are using satellite imagery you can't be guaranteed that you will have a clear and cloud free image so you need to bear that in mind there are multiple different variants of the packaging and I'm not going to go into that you can check other resources for the exact data you want but most of them will include the bands that you will see me using here so let's see let's switch to Kiwis either that the downloaded file is a zip file and I've extracted that to a folder and this is more or less the folder structure for this particular product and depending on the product it will look a bit different maybe so somewhere in this folder structure you will find the raster imagery data and in this case it is sorted into the pixel size for Copernicus or Sentinel data there are 10 metre bands 20 metre bands and 60 metre bands and there are in different wavelengths the ones I'm going to use are all in the ten meter band range and here you have the band two three four and eight you also have some other bands and if you're impatient you can just look for any band that's called TC I that's true color drag and drop it into the canvas and you will have a perfectly usable true color satellite imagery and as I said it's in ten meter range so if we pixel match it in qgs this is the kind of resolution you can expect from this type of image so it won't give you you can you can't read any license plates with it but for example agriculture it's probably good enough unless you are really small plots that you mean to take care of so let's see yeah bands what kind of bands do we have let's go back to Center no - so Santana - has these pants band one is coastal aerosol and I'm not sure exactly how you use that but it's a 60 meter band so it's not really good resolution but it is probably good enough to do some some analysis in the climate sector for instance we have the blue green and red bands and I'm going to use those to start with to create a color image by myself and not used already made one and then I'm going to use band 8 that is near-infrared and that is going to be used to do some vegetation index calculations and they are all ten meter baps so they are all in this folder here so let's add band two three and four and as you see here if we switch to the layers and - black and white and three black and white pan for black and white and that is because the bands are in a certain wave length range and the value of each pixel is the intensity of the light in that wavelength so if it is really bright in the let's see if we make sure I say the right ones band 4 is read then it has a large portion of red in the band but to get a complete picture you need to combine those three and you can do that by going to the raster menu and let's see X miscellaneous build virtual raster and we need to pick our input layers 2 3 & 4 okay the order is important but we can work around that and we want to make sure we have placed each input file into a separate band if we are creating a mosaic with neighboring times of raster data we can remove that but now we want each band to be in each file to be in a separate band and I'll just do it to a temporary file ok that's done now we have a virtual layer and I'll turn the antwon's off that is color but if we compare to the true color it doesn't quite look the same for instance the C is red and that is probably due to the order of the dance so red is band 1 but when I added it we are rated a time in the order band 2 3 & 4 and bound for is the red one and that is the third one in the order that we added them so by flipping red and blue here we should get a more reasonable color scheme it's not perfect yet but just wait in the min/max values we can use different combinations for these layers well if I take mean standard value and apply and we can add a bit more standard deviations like that perhaps can decrease this brightness let's try okay it's still calculating these files tend to be really big so it's still calculating just be patient I'm not patient so there we go now they updated I don't think I'm going to use make a science out of this but you can do a lot of tweaking here and hopefully you can get exactly the image that you want so why should you do this when you already have a true color image well you may not be completely satisfied with a two color image that is why this is really useful to be able to create your own true color image and it isn't harder than creating a virtual layer and applying some tweaks to the bands you have and if you want to save this you can always export it as a file or to a database so that's the true color imagery but I want to use some calculations as well and the thing I'm going to take a look at is something called NDVI or normalized difference vegetation index and this is a way to look at wavelengths in different areas and calculating how the vegetation is how healthy it is or if you are required to make some form of adjustments in your crop yield or your watering or if you need to add nutrients or something like that I'm not an agriculturist so I'm not sure exactly what to call it but I will at least show you how you can create an NDVI imagery with qgs so let's see for this we will create red to green map where red is bad and the green is good the values are kind of important so we will need to make some adjustments for that and the formula we are going to use let's see is here it's near-infrared - the red / near-infrared + the red and that will give us a value between minus 1 to plus 1 and it matters if it is minus or plus so we can just stretch the red and green between the maximum values it really needs to be standardized but more on that later so near-infrared - red near-infrared + red and then we need to add near-infrared and that was band 8 that so we have about 4 Amanda it goes to and to create the near and the NDVI layer we can use the raster calculator and processing toolbox so just type raster calculator and you will get this now I'm not sure but there are some predefined expressions and I actually don't remember if I created this I don't think so so this should probably already be there maybe so I can't guarantee it but I will use it to start with it will bring up a new what is the cool a new window with selectable areas what is selectable layers anyway nearly infrared that was bad 8 so we select band 8 and the red is bad for select plan for okay now we have his layers bound 8 near-infrared - band for the red / I haven't gotten this to work the ampersand so I'm going to use the normal / near-infrared + red and that should be placed in a temporary layer and I think I may need to set a reference layer and I only need to select one of them so I'll select the red layer it says optional but I have had some problem getting it to work without setting that one so let's try there we go done so now we have an output layer that stretches from minus 0.6 to almost 1 and now we need to do something about this it's a single bound it's portrayed as gray and I want it in pseudo color I select a color ramp that goes from red to green if it isn't already selected and to be able to compare data from one day to another you really shouldn't just accept the default interpolation here it should be exact values know one of me you should not have it continuous yes you should what am I talking about you should change the min Max values so it goes from negative 1 to 1 because that's the max range why well if you just stretch it from the lowest to the highest value you could end up with red values for 0 because you don't have any negative values and if you compare a picture with another one where you have negative values that will give the impression that they are similar when they're not if you don't know what I mean so by setting the exact scale that minus 1 that is red and plus 1 that is green and 0 is yellow you can actually compare an NDVI picture from time to time and see the progress if your changes in you're fertilizing or whatever has made any changes okay okay so here we have NDVI and renamed and DV I we see it's some red areas and some green areas and the red areas is supposed to be call it bad and the green areas is good you need to consider that they these calculations is made for vegetation index so when you have an area that is without vegetation it gives you completely useless results so water for instance that that is pretty useless in as an NDVI product but I'm guessing these are fields for growing crops and then at least we could compare those with each other to see which fields are better or in worse shape so here for instance seems to be doing good and this area that probably needs some attention and if you go into smaller scale we see that there is something that is bad better in the center of this field I'm not sure why and then there are darker or rather more to the negative around it so if you are a farmer this I assume this could be really useful maximizing your crop yield yeah that's it that was showing how to create or manage multispectral imagery in qgs without any plugins so if you are doing this on a more professional level or are going to use it extensively and think that this is a bit blunt tool to use there are plugins for this and one of the most talked about is semi automatic classification plug-in it is a really extensive plugin and it requires some other libraries so I'm not sure even sure if it worked for me there right now but you can read up on it on the their home page and they have the YouTube videos to train you in how to use it and there's a user manual etc and it really is extensive you can do a lot of remote sensing as it's called or a multispectral imagery analysis with this plug-in but it will require a few extra libraries and it is not simple to use without training so don't expect to get a lot of out of it unless you actually read the user manual or use the tutorials it really is a great plugin but if you are not going to use it extensively it is good to know that you can do simple stuff like this without it and you don't need any extra libraries or anything to use it in cuties like this so that is all see you next time [Music] [Applause] [Music] [Applause] [Music] you

Channel: Klas Karlsson

Views: 11,820

Keywords: QGIS, NDVI, MultiSpectral, Analysis, Processing

Id: j61X3l6LvxY

Length: 26min 44sec (1604 seconds)

Published: Fri May 24 2019