Site mapping with 6" accuracy

[dougcjohn]

I enjoy getting into maps, orthos, 3D models. I've experimented with multiple products... and the time in processing has been lengthy and rewarding. Intend to continue... also looked at some video options too; new SkyBrowse is gaining a lot of traction... particularly in accident reconstruction or when the on-site needs to be prompt... SkyBrowse is now expanding to night or thermal too. I'm a little hesitant to use due to cost for my low volume needs.

Although for quick projects or limited number of projects, is the sUAV and associated software the best route?

Hadn't previously considered, nor examined in-depth until recent, due to an Insurance Roof estimate after hail storm, I was impressed by the various aerial services (eagleview, snaproof??, etc) the Insurance and Roofing shops utilized. They provide professional prepared reports, apparently close measurements & calculations, and they offer similar service for other forms of ground projects: terrain modeling, construction... can provide reports, detailed diagrams and measurements and/or files to use yourself.

The cost isn't excessive per project, results in 1-2 days.

It's another option to consider when it's not your daily activity or you lack the tools or knowledge to produce quickly.

 

[R Martin]

Based on my experience, the accuracy of the point cloud obtained from an photogrammetric aerial survey using a UAV depends on the following:

1. The resolution of the camera,
2. How high you fly,
3. How fast you fly,
4. How smooth the air is,
5. The contrast of visible points on the surface being measured, and
6. Lighting conditions.

You can get 6” accuracy from a low altitude flight with a 12MP camera. I’ve done it with a DJI Mavic Pro. Any of the DJI UAVs that have a 12MP or better gimbal-mounted camera that can look straight down will work for what you are doing. I suggest you stick with DJI products since most of the available mission planning software supports them, but select and check the mission planning software first to make sure it supports the UAV before buying the UAV.

That is not an accurate statement. You are confusing GSD with accuracy. The resolution of the camera and altitude determine your ground sampling distance. The lower the GSD the more detailed the imagery (ie, the greater the resolution of the imagery).

Accuracy is determined by the quality of the GPS unit involved, and prosumer UAVs do not have an extremely accurate GPS unit (measured in meters) installed in them off the shelf. Without an RTK/PPK solution (to lower the accuracy into the centimeter or better range) to bolster a prosumer UAVs GPS accuracy and/or usage of GCPs that have been accurately set using an accurate GPS unit, your maps are nothing but pretty pictures. With all due respect.

 

[R Martin]

"Making actual maps, including aerial maps using UAVs, for commercial purposes is usually restricted to professional engineers and professional surveyors by most states. "

Er, not sure this statement is accurate. Anyone can engage a UAS operator for purposes of creating a geomap, without state interference. It's done everyday. However, if you are referring to making a map and calling it a "survey" for property lines purposes or claiming one can create maps of "survey accuracy" then you are right; you'd be asking for trouble.

For what the OP is asking for (nothing of survey quality or tied to a datum point) any 'ol P4P with the right processing software, and following your guidelines, will be fine.

You are correct. Surveyors are the only one that can legally survey a property (or someone working directly under a surveyor using his/her stamp). Everyone else is providing a mapping service. Mapping services are not legally binding. Surveys are.

 

[Darcy]

New construction is never present for months or years.

This is a very valid point and is the main reason why we'd like to use drone mapping. I don't know if there is a standard interval for how often any one particular area is updated on Google Earth, but it usually seems to be several years at least.

 

[R Martin]

This is a very valid point and is the main reason why we'd like to use drone mapping. I don't know if there is a standard interval for how often any one particular area is updated on Google Earth, but it usually seems to be several years at least.

It varies on Google's update cycle. Different areas have different priorities. I have some portions that affect us that are four years out of date. We use county pictometry as part of our base map and update that with new tiles that we fly. Not sure of the process required for Google though.

 

[dougcjohn]

This is a very valid point and is the main reason why we'd like to use drone mapping. I don't know if there is a standard interval for how often any one particular area is updated on Google Earth, but it usually seems to be several years at least.

My small MidWest City, despite being State's Captial. The Google maps are over 4 yrs... several sites built in last 4-5 yrs no on Google. My office parking lot shows a car I drove over 4 yrs ago.

Not so bad when Streets & Blocks remain unchanged, but a problem when new / recent developments... that Google shows as no streets and woods.

 

[R.Perry]

Ah, what the heck, what is a foot or two among friends.

 

[R Martin]

Ah, what the heck, what is a foot or two among friends.

On a small job chump change. On a larger job you are bidding, the difference between turning a profit and losing your rear.

 

[Earthman]

That is not an accurate statement. You are confusing GSD with accuracy. The resolution of the camera and altitude determine your ground sampling distance. The lower the GSD the more detailed the imagery (ie, the greater the resolution of the imagery).

Accuracy is determined by the quality of the GPS unit involved, and prosumer UAVs do not have an extremely accurate GPS unit (measured in meters) installed in them off the shelf. Without an RTK/PPK solution (to lower the accuracy into the centimeter or better range) to bolster a prosumer UAVs GPS accuracy and/or usage of GCPs that have been accurately set using an accurate GPS unit, your maps are nothing but pretty pictures. With all due respect.

I appreciate what you are saying and understand the connection between photo resolution and GSD, but I’m talking about the overall accuracy of the aerial map prepared using only photogrammetric techniques, which doesn’t rely on GPS. We made accurate aerial maps using only aerial photos long before there was GPS. So, I disagree that the accuracy of a photogrammetric survey has anything to do with the quality of a GPS receiver.

I believe that you can get 6” contours with an accuracy of +/- one contour interval using aerial photos from a 12MP camera. How accurately those contours are tied to a standard datum/reference frame Is another issue that may depend on the quality of a GPS and/or GCPs.

 

[Darcy]

However you collect your aerial data, will not have an “aerial map” that includes surface features such as edges of pavement, sidewalks, mulched areas, landscape walls, etc. You will have to trace and add these manually in CADD using the orthophoto, etc. This is time consuming and expensive. So if you need this level of detail, you will need to become a mapmaker.

Earthman, could you elaborate on this, please? You're saying an orthophoto is not the same as an aerial map showing surface features? I thought that's what an orthophoto is. But, I've never had the opportunity to create an orthophoto, so maybe I'm not understanding what their final product is.

 

[R Martin]

Earthman, could you elaborate on this, please? You're saying an orthophoto is not the same as an aerial map showing surface features? I thought that's what an orthophoto is. But, I've never had the opportunity to create an orthophoto, so maybe I'm not understanding what their final product is.

An orthophoto is just that, a photo. You can roughly scale things from it if you have a know point of reference. A more accurate way to use the information is to digitize features from it using either CAD or GIS software. In essence, you are recreating reality digitally so that you can measure or interact with the information more accurately.
To do so you need at a minimum a basic understanding of the software in order to use it. There are freeware and shareware versions available of both with more limited functions. I have Autodesk Cad and ESRI ArcGIS and both are really expensive.

 

[Earthman]

Actually, an orthophoto is an “accurate” representation of what is in the photo, so you can measure between objects visible in the photo. I only put the word accurate in quotes because everything has errors in it and you should understand the affect of the errors on your use of the data. For many non-critical uses, the errors in an ortho may not be important.

The ortho may be considered a simple map, but it’s not going to include all the features that one may need to be useful. Missing features may include ground surface contours, property lines, fence lines, benchmarks, layout and control point locations, edges of pavement, sidewalks, surface features of subsurface utilities, subsurface utility size and depth, building corners and overhangs, cracks in pavements, and/or any information that is important to the enduser of the map. All of these kind of features need to be added to the drawing/map using CADD, etc. Some of the features can be traced into the drawing/map from the ortho, but others need to be surveyed.

The orthophoto is just a photograph of the ground and what obscures the ground (tall weeds, tree canopies, etc.), so it’s not that useful by itself. It’s one tool that can be used to help create a useful map.

 

[Dave Pitman]

An orthophoto is just that, a photo. You can roughly scale things from it if you have a know point of reference.

Actually, an orthophoto is an “accurate” representation of what is in the photo, so you can measure between objects visible in the photo. I only put the word accurate in quotes because everything has errors in it and you should understand the affect of the errors on your use of the data. For many non-critical uses, the errors in an ortho may not be important.

Same same.

An orthophoto can be geographically very accurate or very inaccurate, and everything in between.
An "inaccurate" orthophoto, by survey standards, may be very useful for say, a landscaper. ?
And, very accurate orthophotos are often dismissed by skeptical surveyors too!

 

[Darcy]

A more accurate way to use the information is to digitize features from it using either CAD or GIS software. In essence, you are recreating reality digitally so that you can measure or interact with the information more accurately.

Can an orthomap with only relative accuracy, no defined global accuracy, be imported & manipulated in GIS software? I have not used any GIS software, so I have no experience in this regard.

 

[R Martin]

Can an orthomap with only relative accuracy, no defined global accuracy, be imported & manipulated in GIS software? I have not used any GIS software, so I have no experience in this regard.

Sure. You can import the ortho and georeference it where it snaps onto an existing base map. Assuming there is a spatial reference associated with the ortho you can transform that into another spatial reference with the tools provided. It's an easy process. If you import an ortho that is not in the same spatial reference as your working document, Arc will always prompt you before to transform the coordinate system before the import proceeds. You can always opt to do nothing and proceed as well.

As a rule, I always export the ortho in the coordinate system I am using in Arc from Pix4D so when I do import the image, its drops where it is supposed to be. The default in Pix4D is WGS84. My transformation is to NAD83 2011 Texas State Plane 4202.

 

[Earthman]

If you don’t have GIS or Pix4D, you can use other, lower cost ,software to georeference the orthophoto or any map/figure. I’ve used TopoGrafix to georeference all kinds of scanned maps for use in GoogleEarth, CAD and GIS. You need four features near the corners of the map that you can find in both the figure and GoogleEarth. Depending on the scale and size of the map or orthophoto, these can be road intersections, building corners, etc. once you get the XYZ coordinates of the four features from Google Earth. use TopoGrafix to add the points to the figure that you are georeferencing. Then you can open the figure in CAD or GIS and it will open in the right location to the level of accuracy of the coordinates of the features you picked.

There are also low cost alternatives for CAD and GIS software. Google “free CAD” or “free GIS.” Also search for open source programs,

 

[Darcy]

Thanks for all the good advice shared on this topic. I’ve learned quite a bit, since I’m totally new to this field of drone mapping. The process that’s forming in my mind for our application goes like this:
1. We’ll lay out several physical targets of a known size on each site before flying a mission.
2. Using MME (to keep initial costs low), we’ll create a stitched orthophoto from our images
3. Our landscape design software allows us to import an overlay image and resize it by referencing a known measurement on the image. So we’ll import our orthophoto and adjust it according to the targets we set on the site. Then we can trace over the areas/distances we need to measure.
4. Alternatively, Step 3 could possibly be done in MME, thus eliminating the step of importing into our design software for some simple projects.

I realize this method won’t be as accurate as some methods suggested, but I feel we need to balance accuracy, acquisition/processing cost & time. We’ll have to take a few manual measurements to determine our accuracy on our first several projects until we’re comfortable our process is actually providing sufficient accuracy.
Worst-case scenario is we'll find out this process simply won't provide consistent accuracy, but at least we won't have sunk a large chunk of money into it yet. We'll have some experience under our belts and be able to re-evaluate whether it's worth pursuing further.

 

[Earthman]

I think you are on the right track, but in terms of scaling the ortho photo, you would be better off knowing the horizontal distance(s) between points, objects or features that appear in the ortho. This might be the distance between the corners of buildings or the width of a road, or the distance between targets. The longer the distance(s), the less error in your scaling. You might want distances in two orthogonal directions (north/south and east/west or similar). Although you can do it, you don’t want to try to scale an ortho of a relatively large area using the dimensions of a relatively small target.

In a regular aerial survey workflow, you enter the surveyed XYZ coordinates of the GCPs (targets), and the software distorts the ortho and point cloud to fit the surveyed points.

I suggest that you use the Maps Made EZ App (MME) to plan your mission, program your drone with the mission data, and to post-process your photos. They will provide an ortho and point cloud for acres of area for free. You don’t need to worry about scaling or using GCPs to get an uncorrected survey with elevations relative to an arbitrary datum. This may be all you need for the mapping part of your workflow.

Do a trial survey in a field and let us know how it works out for you.

 

[G Saunders]

Hi, I worked for a landscaper a long time and used DraganFlyers to pop up over a site and get video, then would pull frames from that and manually stitch them together for the background on a CAD drawing. They were accurate enough to be helpful. The DraganFlyers were flimsy, hard to fly, and video was more like CCTV than a 12 MegaPixel image.

I recently got this yard, can't call it a map or a survey, by hand-flying under the trees at about 75 feet, no ground control points, and sent the pics off to DroneDepoloy. I've done it a few times and hope to get better. You can see it's off a bit, but is this 'good enough' for what you're doing? I'm not working with landscaping anymore, but one like this was used by a tent company that put up a tent that pretty much covered the whole yard for a wedding. You can see the patterns in the grass between the tables in this shot, taken after the wedding and the tent was taken down. The owner was chagrined to see how far out of whack he got the fence around that fire pit.

Since most every yard I expect to be visiting around here is under trees I'd seldom get a chance to let the drone fly over the trees and get what I need, so I'm practicing this. I'm hoping Skydio 2 will be able to automate the task with its advanced obstacle avoidance, meanwhile I fly these by hand, watch the FPV and click the shutter button for about 2/3 overlap...

 

[Darcy]

I think you are on the right track, but in terms of scaling the ortho photo, you would be better off knowing the horizontal distance(s) between points, objects or features that appear in the ortho. This might be the distance between the corners of buildings or the width of a road, or the distance between targets. The longer the distance(s), the less error in your scaling. You might want distances in two orthogonal directions (north/south and east/west or similar). Although you can do it, you don’t want to try to scale an ortho of a relatively large area using the dimensions of a relatively small target.

In a regular aerial survey workflow, you enter the surveyed XYZ coordinates of the GCPs (targets), and the software distorts the ortho and point cloud to fit the surveyed points.

I suggest that you use the Maps Made EZ App (MME) to plan your mission, program your drone with the mission data, and to post-process your photos. They will provide an ortho and point cloud for acres of area for free. You don’t need to worry about scaling or using GCPs to get an uncorrected survey with elevations relative to an arbitrary datum. This may be all you need for the mapping part of your workflow.

Do a trial survey in a field and let us know how it works out for you.

Thanks for the feedback! Yes, that's great advice to use longer distances for verification measurements. We might have to invest in a Leica Distance Measurer for taking those.