Agisoft Metashape Tutorial - Full Processing from photos to dense cloud - with Ground Control Points

hello this is a full tutorial for address on photoscan professional I wanted to bring you guys to the entire process at least to generating a dense point cloud so basically I've had to stop photoscan open with no data first up is workflow you're going to add some photos so I have an existing project these points are 80 percent overlap 80 percent side lab and these are the photos I'm going to start with so you can see the photos are toggled on and the top menu bar with the photo icon so you can toggle them on and off if you want to see them and the next step we're going to do is actually reproject these photos from wgs84 latitude longitude to state plane coordinates so go ahead and hit the convert button not the Settings button so the settings button is actually for assigning a coordinate system and I have another video tutorial on that if you're interested I actually go to convert this is going to convert these coordinates that are given from the EXIF tags I'm gonna go ahead and use Wisconsin Central that's epsg 29:29 go ahead and select that so once you get that selected you'll see that actually projects these into that state plane coordinate system zone which is going to be very helpful when we bring in our ground control because they're basically going to line up pretty close to the ground control points so once you get that converted the coordinates converted you can go ahead and align the points so this takes a little bit of time pause fast-forward that but first thing I'll go over is the key point limit to trial and error I've settled on 240 thousand for the key point and the tie points I actually don't limit those at all so placing a zero and that will actually not limit the type points I've had the best luck with this especially with high megapixel cameras if you get a 30-plus megapixel camera you're going to need more more key points and tie points enabled so go ahead and hit OK it'll start processing I'm going to fast forward this for you guys so you don't have to watch the entire statusbar though I like to look at the details and kind of see what's going on video editing I'm gonna be back with you guys in real time real time speed shortly here we are so this has generated what's known as a sparse cloud for for our project and we could kind of zoom around you're going to see there's some points but it's not a dense cloud but it means it's just kind of the first step towards the dense cloud so the next step is actually to go ahead and import some ground control points so to do that you go to import I have a DCP file right and data city slip there we go do you see peas and here we are so I'm gonna go ahead and select this one and now you see the previews in the bottom and I believe this is an XYZ file so I'm gonna go ahead and change these labels to be call for change the easting to be column one change the noir thing to be column two which it is and then altitude or elevation should become 3 so that's XY and z and i can confirm this by going up and looking at the projected coordinates on there so here I have the northern is 23 or sorry 253 easting is 248 so yeah it's actually not it's actually in northern easting in elevation so northern column 1 Dec column 2 now it matches up northern northing okay good so I can go ahead and create those and being I don't have in the project I'm going to hit yes to all so there we go we have our points created and here's the tip guys if you ever want to get back to the top view you can go ahead and hit the 7 key so if you hit the 7 key it'll actually bring you to a top view kind of north up and then if you go somewhere else and kind of pan around and you want to go back there you can just hit the 7 to get back there that's a good tip and another one if you look up on the top left of the model of the window you'll see I'm on a perspective view and then pressing five actually toggles with orthographic for this process and a lot of the aligning I recommend orthographic perspective is a cool view it's more 3d but it's not as practical for things like magic point so I'll leave it at north of graphic and now I'm gonna head up here to two markers detect markers and and it'll actually detect these cross markers that I have on the ground and I've found good luck with 50 and 10 I'm not sure if their default but you can adjust those for your given data set so now I'm going to go ahead and let the computer detect markers so it's actually looking through all the photos and trying to recognize that cross shape which is which is a ground control point pattern and when it's done I'll go through a couple photos and show you what it's it's recognizing it does a really awesome job so now you can see there's actually duplicate points here and I'm gonna go through a process of actually renaming those generated points to to our existing ground control point numbers and kind of bear with this video I think you'll see why I'm doing this here at the at the end of the swimmin arena at that point one and then this one here another tip for you guys is toggling that selection back and forth is the spacebar so that's gonna be helpful 2.25 it's actually 0.2 there you go I'm going to go up here select that one change at 2.3 selected their backs and so here go ahead grab those yep so change the I point to point for these auto-generated ones there you go point five there's another one point six and then point seven seems to be kind of straight I think it might be a control point outside of the project boundary so I'm gonna go ahead and wipe that point seven out we don't need that point seven so go ahead just right-click and remove it go ahead and remove it now you can see we have duplicate point numbers so I'm going to sort these by easting now all of these on the bottom were imported already from our file so I'm going to delete those now all this up to these auto-generated ones the problem is there's no northerns these things are elevations associated with those so now we're going to re-import that crowd control point file now you could see that's the same thing we did now it didn't ask to create new ones because they aren't existence and now it just assigned the northings and eastings and elevations Telos now can zoom in now this is what it's recognizing automatically so I think edges off does a great job recognizing these non coding targets I haven't heard of people having very good luck with the coated targets at at drone elevations may be for an archaeological site or something they work but for for the drone application these non-coated targets seem to work pretty great I'd like to hear your feedback also so another thing I recommend is going ahead and and selecting this little yellow triangle that shows your accuracies so you could see some of these are you know a little bit up there but if you go ahead and update those and then optimize here and I typically use adaptive model fitting and then estimate type point conversions so I'll let this run this takes a few minutes and ultimately what this is doing is doing a kind of a constrained adjustment on those points so you'll end up with should be much better estimated error accuracies for your control points a lot of times I'll open that details just to make sure that the program is still running and that we have good progress being made so this generally goes kind of quick you know it seems to under estimate the processing ability of the computer at first and and it does seem to speed up pretty well so again guys this is kind of a lull in the video so I'll take this time to say if you're finding value in these videos just like the videos and subscribe to the channel so now you can see those accuracies are much better you're down in the neighborhood of you know one hundred to hundreds of a foot which is about a eighth to a quarter inch of estimated accuracy so it's it's really constraining these and tighten up to the ground control very accurately so the next step here the final one of this tutorial let's go ahead and build a dense cloud so I generally don't recommend using ultra high but in depth filtering if you guys have any really good advice there on depth filtering feel free to comment below and I'll typically calculate point colors I don't know how much time it adds but it really does help the view so imma go ahead and kind of increase the speed of this video and actually cut out a little bit of it just till we can get to the to the end and take a look at that dense cloud there is an option to do network processing as well actually on a hosted computer by add to soft I have another video outlining that procedure I encourage you guys to use it it it really does save a lot of time especially if you don't have a really big graphics card okay here's the final result so this is the dense cloud processed with ground control and I'll kind of zoom in here and take you guys around these piles and then you'll get a pretty good idea of the density and the quality of this this point cloud it's it's really good I've noticed that when they switch from photo scan to meta shape it's really improved the quality of these these point clouds very good so that's as far as this video is going to go guys I have other videos that are going to guide you guys through the process of building a mesh and a de M if that's of interest to you please check those videos out and little zoom in here to see the the quality that dense cloud and now it creates the the colors that show that target thanks guys

Channel: SkySurv LLC

Views: 52,380

Keywords: drone, dji, photogrammetry, GCPs, Agisoft, Photoscan, Metashape, 3d

Id: EZfahR49lxA

Length: 12min 1sec (721 seconds)

Published: Wed Jul 31 2019