What is the difference between using a drone to take pictures to generate topo versus "standard" photogrammetry? There is a pretty heated argument in my office over the last few weeks and I'd like to get some "outside" feed back. Thanks.
??? Aside from the obvious differences in altitude, speed, and camera, there is no difference. Photogrammetry is photogrammetry.
W.W.S., post: 412279, member: 11419 wrote: What is the difference between using a drone to take pictures to generate topo versus "standard" photogrammetry? There is a pretty heated argument in my office over the last few weeks and I'd like to get some "outside" feed back. Thanks.
What are you arguing about? It is photogrammetry just with a smaller aerial platform and camera format.
SHG
They're not the same, I can't even wrap might mind around people saying they're the same. One is people manually measuring points with their eyes in a stereo compilation system, the other is software looking for the highest confidence level.
What's the difference between you sighting a backsight and a robotic total station turning an angle and measuring a distance?
Traditional photogrammetry relies on humans to interpret the imagery, the new technology is 100% statistics based. If you think those techniques are equivalent then you've never sat in a compilers seat and looked at automated data.
You'll be getting a very simplistic survey if you just accept what comes out of the software. Anything much more than a bare earth surface needs a lot of detailed human study to extract a result approaching a traditional photogrammetric or conventional land survey. (And quite a lot of check fieldwork).
It depends upon what use you have for the data - measuring crop growth is probably fine when fully automated, so is quarry excavation as long as there aren't vehicles around (yes you can filter, but you might also remove heaps of rock). On that point, has anybody a successful solution to the tendency for Photoscan to smooth off the extracted point cloud along the intersection between horizontal and vertical faces? I think the same problem arises in Pix4D as well.
chris mills, post: 412620, member: 6244 wrote: You'll be getting a very simplistic survey if you just accept what comes out of the software. Anything much more than a bare earth surface needs a lot of detailed human study to extract a result approaching a traditional photogrammetric or conventional land survey. (And quite a lot of check fieldwork).
It depends upon what use you have for the data - measuring crop growth is probably fine when fully automated, so is quarry excavation as long as there aren't vehicles around (yes you can filter, but you might also remove heaps of rock). On that point, has anybody a successful solution to the tendency for Photoscan to smooth off the extracted point cloud along the intersection between horizontal and vertical faces? I think the same problem arises in Pix4D as well.
Chris and Geopro, I TOTALLY agree with what you are saying. Close Range Photogrammetry is a disaster waiting to happen when used by someone who automatically trusts the results. The problems are easily demonstrated, and those who are not aware of them will one day be absolutely shocked (and hauled away). Often the problems are well hidden.
If one thinks GPS multipath causes risk, close-range photogrammetry has exponential risk when used without a lot of experience, knowledge of how it works, and redundancy checking. At some of the many seminars I have presented across the country for CEU credits, I have had many people come up to me saying things such as with a gopro and a drone they surveyed 40 acre topo and found nothing more than 0.1 ft. off. Statistics might have shown it, but I don't believe it. I have an antique clock in my office that does not run. But it is exactly right every day at 2:20 pm, even as accurate as an atomic clock for a brief moment. While close-range photogrammetry is remarkably useful, it is not as simple as conventional orthophotogrammetry and has many more variables that must be understood to use it professionally. It irks me that GIS people are pushing buttons and practicing surveying without a license and getting away with it--ultimately at client risk or danger of loss, which is why surveying is regulated in the first place.
Frank Willis, post: 412623, member: 472 wrote: Chris and Geopro, I TOTALLY agree with what you are saying. Close Range Photogrammetry is a disaster waiting to happen when used by someone who automatically trusts the results. The problems are easily demonstrated, and those who are not aware of them will one day be absolutely shocked (and hauled away). Often the problems are well hidden.
If one thinks GPS multipath causes risk, close-range photogrammetry has exponential risk when used without a lot of experience, knowledge of how it works, and redundancy checking. At some of the many seminars I have presented across the country for CEU credits, I have had many people come up to me saying things such as with a gopro and a drone they surveyed 40 acre topo and found nothing more than 0.1 ft. off. Statistics might have shown it, but I don't believe it. I have an antique clock in my office that does not run. But it is exactly right every day at 2:20 pm, even as accurate as an atomic clock for a brief moment. While close-range photogrammetry is remarkably useful, it is not as simple as conventional orthophotogrammetry and has many more variables that must be understood to use it professionally. It irks me that GIS people are pushing buttons and practicing surveying without a license and getting away with it--ultimately at client risk or danger of loss, which is why surveying is regulated in the first place.
Have you done a lot of cross checking? It seems to me that the idea of the WalMart camera put into a drone and a few ground points could cause a bit of an issue.
I know that all the years we did photo-control for some of the more senior photogrammetry companies there was a spread between them as far as accuracy is concerned. We did a lot of checking and errors sometimes weren't too hard to find. So a drone, a couple of ground points and inexperienced users, well............
There's a big difference between throwing a "Wal-Mart camera" on a Phantom 3 and what we do with our UX5-HP. In addition to GCPs and ground truth points, the software provides ample information about the quality of the data. As far as relying on statistics, anyone who uses GPS and / or least squares is relying on statistics. UAS is just one more tool in the toolbox; used carelessly or incorrectly there's potential for bad data. Fortunately I'm neither careless nor negligent in learning to properly use my tools and software.
Lee D, post: 412644, member: 7971 wrote: There's a big difference between throwing a "Wal-Mart camera" on a Phantom 3 and what we do with our UX5-HP. In addition to GCPs and ground truth points, the software provides ample information about the quality of the data. As far as relying on statistics, anyone who uses GPS and / or least squares is relying on statistics. UAS is just one more tool in the toolbox; used carelessly or incorrectly there's potential for bad data. Fortunately I'm neither careless nor negligent in learning to properly use my tools and software.
True but you don't need to spend a fortune to get good results. An Inspire 2 or even a P4 Pro will do the job for most small to mid-sized projects with their 20mp cameras. The P4 Pro has a small sensor but in the right conditions and for the right project it can be the right tool (sometimes your tape measure is the right tool over your total station). The camera on the Inspire 2 is pretty good quality these days and the camera is what really matters. You can build and fly a DJI s1000 or M600 with a Sony a7 camera for $10-$15k. Pair that with good ground control and you will have the same results as some of the more expensive options (probably with more flight control flexibility). We work with some people in the film industry who all use DJI systems. Sometimes they are using an Inspire other times they are using a heavy lift m600 or older s1000 with a variety of cameras.
For those looking for more info on the process take a look here [MEDIA=youtube]NGdZ8O2cWks[/MEDIA]
We have compared our sUAS photogrammetry to projects we have had flown traditionally to ground topo to 1 meter DEM (LiDAR) data and found very little difference if everything is done correctly (that is the Key right now!). The real benefits are time and cost savings. We also use close range photgrammetry for ground based projects. We recently did a ground based test (using a Sony a6000) of a site that we had previously done ground topo on and found the results to be very good. The vast majority of point cloud points checked within hundredths of our ground shots. The process needs some refining (took forever to process for a variety of reasons) but both us as the users and the software itself are only getting better.
Tom
Thank you all for your thoughts and replies.
We contract out photogrammetry and require a Certified Photogrammetrist. We supply the RMSE values for the mapping product
We also require:
The USGS camera calibration report.
A triangulation report for pass-points/tie-points, ground control points, residuals, etc.
The standard 2014 ASPRS cert for "This data set was tested...' or "This data set was compiled..."
or
The standard 1990 ASPRS cert "This map was checked..." or "This map was compiled..."
These are to be on the face of the plan and in a letter signed by the CP.
So as a PLS is it acceptable to go out and map using a drone and a 50 megapixel Cannon SureShot (can I get a calibraiton report? and does it matter??) and Pix4D as long as I do the standard bare earth and vegetated earth check shots to test the data? To me it feels like crossing into CP territory. To a few others in the office it is the next tool in the toolbox.
Of course we contract out aerial and mobile lidar, but use terrestrial lidar....
I am not the decision maker, but just trying to best educate myself.
It seems to me that there are 2 separate issues at hand here.
- Traditional 12" format photos taken from a manned aircraft at altitude vs. UAV/consumer grade camera photos
- Stereo plotter map production vs. programmatic point cloud based mapping.
And as with any mapping system there is no replacement for getting out in the field and collecting some "confidence points" to test the model, however it may be produced, and whatever the statistical report says.
As for the comparison to least squares - with any LS adjustment one should always examine the residuals to determine if any raw measurements differ from the adjustment results by more than the comfort level. If a large number of measurements were taken a few outliers will not shift the final statistics enough to cause failure but may still be significant.
I think that an easy answer is Accuracy. Drones fly at a much lower height and are capable of much more accurate results (when used correctly) than standard photogrammetry fly overs.
Surveying and mapping laws DO NOT typically legislate methods. The accuracy argument could be used for/against collecting positional data from a "mapping grade" reciever vs. collecting that data with a "survey grade" reciever. Believe me I have heard the argument that it is only good to a meter so we don't need a surveyor. Show me in any state surveying law where positional accuracy is required on a utility pole for instance. Yet everybody seems to think if it gets located to the millimeter then somehow you cross the magical line into surveying. As equipment and methods get better and lines blur, I don't think the accuracy of a product can dictate whether it falls under one profession or another.
I maintain rather from a drone or a manned aircraft, the knowledge base needed to accomplish the mapping and certify to the results is largely the same rather you call yourself a CP or a PLS. The bottom line is you need to know enough to know if you have garbage or a good product from any method you use.
SHG