Agisoft Photoscan Professional Tutorial: Creating a DEM without GCPs

Video Statistics and Information

Video
Captions Word Cloud
Reddit Comments
Captions
welcome to an edgy soft photo Santa tutorial we are going to add photos to a workspace and build a digital elevation model a mesh texture our mesh and all of the workflow that goes with that so the first thing to do in the software is add photos you can do that under workflow add photos and navigate on your computer to where your photos are stored which I've already done select your photos and add them using the Open button so this is the model space it shows you where your photos are aligned and this is how you can turn on and off the cameras the cameras are shown showing the exif data from the UAV which is the GPS locations of each photograph and what the software is going to do is it's going to align like points from successive photos so this is the workspace where you can see what you brought in so here's my photos for example and the next thing I'm going to do now that I have these photos is I'm going to align them so again under workflow go down to align photos we're going to use medium accuracy on this because for the nature of the demo I want to get this done quickly these settings are all fine make yours look like this while you're doing it and go ahead and let it run and this step can take 10 to 20 minutes depending on how powerful your computer is etc etc so excuse me after you align your photos you should get something like this which is your point cloud so the software creates a point cloud from like points like we mentioned I'm gonna turn off the camera views because they get in the way make it a little difficult to do the next step what we're gonna do is we're gonna refine this point cloud and generate a dense point cloud using this data so you know the first refinement we'll do this here under module gradual selection there's a couple different selection tools here we're gonna use the reconstruction uncertainty tool and what this does is it selects points based on the reconstruction uncertainty so I'm gonna go to about 40 to get all of these outlying points that came from the backgrounds of the photographs that we don't really need for the area that we're working with so I've got those points selected and I'm going to delete them you can use this or the Delete key on your keyboard another way to get rid of points that works pretty well in detail there is so easy I'm gonna use the freeform selection and select the points like this so again delete on my keyboard and delete those oops delete those and one thing you'll notice is after you've made a point cloud this region here has been automatically created by the software and you can refine that region to only include the area of interest for your project which I'll show you how to do right now so now I think I got all the points I want now I'm gonna find this region oops I accidentally grabbed some points so a quick way to select those let's use this tool again and have slider all the way backs and then just hit cancel in it you select with those and I'll click the navigation tool to rotate in my model space I've got this about where I want it and I'm going to adjust this rubian bars so that the software only creates the point cloud in the region that I'm specifying so first things first I'll resize the region and the dark red on the bottom represents the ground so keep that in mind all right that looks good and the other options are to move the region and rotate the region and if you don't like what you did you can always start over I'm gonna go back to the navigation tool and take a look at it I can get it a little all right I think that's pretty good for what we're doing one more small selection here to remove and next up we're gonna build our dense point cloud so in the workflow go to build dense point cloud for this tutorial we're gonna do this on medium and this step usually takes a while the depth filtering is something that you'll vary based on the terrain of what you're modeling so for a model like this that has a lot of elevation change we want to be aggressive if you're doing an open field like a soccer field or something mild is fine and if you're doing a 3d model something with a lot of change in shape you all definitely want to be on aggressive so we're gonna run that and usually this one will take 10 to 20 minutes depending on your computer but like any good cooking demonstration I already have it done in another in another example so here's what a dense point cloud looks like after you finish it and you can see I've got some holes right here where the software didn't quite pick up what we wanted to do and unfortunately that's just the way it goes if I would have ran the model at a higher setting than medium it would have probably filled those in a little bit better so from our dense point cloud we are now going to build a 3d mesh and the way to do that is pretty simple with this software you just go back into the workflow and select build mesh so we're gonna do a height field and we want to use the dense cloud we created as the source data and the face count high is fine you can even get away with going a little higher if you want this should run pretty quickly interpolation is fine and we're gonna go ahead and create the mesh and keep in mind these holes right here watch what happens to them once we build this mesh compared to some of the other points that are a lot better defined and so we finished building the mesh if it hasn't shown up for you if you don't notice any visual changes in the workspace you can see the 3d model that popped up and go ahead and double-click on that and here's our mesh which is basically just a 3d model created from the dense point cloud so it connected and interpolated different points and matched textures as much as possible so in some areas you can see things like the cars don't quite look like cars just yet and again like I mentioned this area where we had holes over here filled in pretty pretty poorly relative to the rest of the surfaces so one other thing we're gonna do in order to make this look a little better for visual and like visual analysis more than anything is we're gonna build a texture let me show you where that is one more time it's under the workflow build texture generic and go ahead and do that this would take a little bit of time the only thing that should take a really long time is the dense point cloud everything else should be under ten minutes based on the settings that we're using now the texture is finished and as you can tell when we get in it a little closer detail things look a lot better and hopefully that didn't take too long to run as well the last thing we're gonna create is a digital elevation model and the way you do that is the way we do most of the things you go under the workflow you were to build de m and you're gonna want to make the settings look like this use the dense cloud not the sparse cloud or the mesh and go ahead and hit OK so your your DM should show up in your workspace and look something like this there's a few options it appears and the last thing we want to do is we want to do a little bit of analysis on our elevation model and then export the results so the first thing we're going to do is we're going to map a profile right here when you do that you go to model draw a polyline it's also right here and you left-click to make your polyline and we're gonna go right through this right here and you right click to end the command and then I made 5 points so if you right-click on one of your points and go to measure then the software will calculate the latitudes longitudes and altitudes of the points that you made as well as the perimeter or the length and then under the profile tab you can see the general shape of the digital elevation model along your polyline and is also the length here we're gonna save that there's multiple file types you can save your profile under I'm gonna make a JPEG and it looks like I already saved one like a good cooking demo then I'll put that away next I'm gonna make a polygon to measure a volume and I'm just gonna measure the volume of the seating area again you left-click to make a point and you right click to end your polygon and then right click on one of your points and select measure and this will show you again a lot of T's longitudes and altitudes of your points perimeter of your shape here's the profile of the shape and about right there it looks like I picked up something a little funny along here where I a little too far outside of this flat area and the nice thing about polygons is you can look at a volume with a polygon so there's three options a best-fit plane a mean level and a custom level best fit fits the points as best as possible on top of the digital elevation model mean level averages all of the points that you selected and custom you can just add an arbitrary elevation so for example 1500 is well below our surface which you can tell is that 2000 but on the other hand 15,000 is well above and so I'm going to stick with the best fit just for our purposes and it gives me this nice volume measurement that I can report back to whoever needs me to perform some analysis so now that I have that I'm going to export my de m which is right here in the workspace you right-click on it and you go down to export de m and we're going to use this first file type and I'm gonna do a raster transformation because I just want this as a JPEG so in order to do that you set your raster transformation to palette and you hit export and when you go to save this you have a few file types I go with jpg and I name it diem and I save this and that's all we're going to do in this demo we've processed some basic photograph data we made a few different results from that data and we've saved some analyses we did so thanks
Info
Channel: Ross Hunnicutt
Views: 31,043
Rating: 4.9163761 out of 5
Keywords: drone, photogrammetry, Agisoft, Agisoft Tutorial, UAV, mesh, DEM, texture, GeoSurveying, Photogrammetry Data, drone photo tutorial, Photoscan, Agisoft Photoscan Professional, how to process uav photos
Id: McMGyo98Ffw
Channel Id: undefined
Length: 13min 59sec (839 seconds)
Published: Mon Dec 03 2018
Related Videos
Note
Please note that this website is currently a work in progress! Lots of interesting data and statistics to come.