My First Drone Photogrammetry Project

Bow Tie Surveyor, post: 420286, member: 6939 wrote:
I processed the Orthophoto and Point Cloud through AgiSoft Photoscan Pro.

How much did the software you used, cost?

rfc, post: 420296, member: 8882 wrote: How much did the software you used, cost?

Using a trial version, but AgiSoft Photoscan Pro is $3,499. It's not quite as slick and polished as Pix4D, but it seems to get the job done.

I've recently produced similar results on a larger scale. It was a 44 acre kids summer camp in a very rural area with moderate tree cover. It was a large scale proof of concept job so we did the whole thing conventional as well as with the drone using our conventional control as the GCP for the aerial mapping (~29 CP's). Drone took about 1,200 pictures. I processed the photo's through Pix4D then classified the cloud into a surface using Cyclone, Pix4D and 3DReshaper. Cyclone & 3DReshaper produced the best model results with 3DReshaper being hands down the fastest option. It took about 45 minutes to classify the cloud to a surface model using 3DReshaper vs. a little less than a day using Cyclone and classifying it manually. Areas where the ground was visible we achieved +/- 0.10' vs. the conventional ground topo; grassy areas it was +0.5' which was pretty much expected. The 3D model of the site Pix4D generated was the real crowd pleaser though. All in all I was impressed with the results and it confirmed my expectations of where it will yield reliable results and where it won't.

Cameron Watson PLS, post: 420302, member: 11407 wrote: I've recently produced similar results on a larger scale. It was a 44 acre kids summer camp in a very rural area with moderate tree cover. It was a large scale proof of concept job so we did the whole thing conventional as well as with the drone using our conventional control as the GCP for the aerial mapping (~29 CP's). Drone took about 1,200 pictures. I processed the photo's through Pix4D then classified the cloud into a surface using Cyclone, Pix4D and 3DReshaper. Cyclone & 3DReshaper produced the best model results with 3DReshaper being hands down the fastest option. It took about 45 minutes to classify the cloud to a surface model using 3DReshaper vs. a little less than a day using Cyclone and classifying it manually. Areas where the ground was visible we achieved +/- 0.10' vs. the conventional ground topo; grassy areas it was +0.5' which was pretty much expected. The 3D model of the site Pix4D generated was the real crowd pleaser though. All in all I was impressed with the results and it confirmed my expectations of where it will yield reliable results and where it won't.

I was surprised that my grassy areas were just about as good as my hard (concrete and pavement) shots. Although the grass was not very thick and they did mow the property a few days prior which helped.

How much do those point cloud processors cost (3DReshaper and Cyclone)?

Bow Tie Surveyor, post: 420322, member: 6939 wrote: I was surprised that my grassy areas were just about as good as my hard (concrete and pavement) shots. Although the grass was not very thick and they did mow the property a few days prior which helped.

How much do those point cloud processors cost (3DReshaper and Cyclone)?

The grass I was referring to was tall native grass. There were small areas of lawn/sod type grass where we also got very decent results.

Cyclone I own as I bundled it in with a larger equipment purchase including an MS60/CS20. The software cost was $8,500 if I remember correctly. 3DReshaper was a demo/trial version. I haven't gone the next step of getting a quote but the application seems to be evolving quickly. They're a HEXAGON owned company and I see on their website they have a plugin for C3D now which I'm pretty curious about. The process of ground surface and building extraction was really intuitive.

I commend you and thank you for posting this here.

Bow Tie Surveyor, post: 420300, member: 6939 wrote: Using a trial version, but AgiSoft Photoscan Pro is $3,499. It's not quite as slick and polished as Pix4D, but it seems to get the job done.

For reasons to be revealed soon, I would recommend looking at PixElement.

This whole thread is very interesting to me. The equipment and software is always an expensive investment, sounds kind of doable for the hard costs but can anybody inform me about what the training costs are for operation of the drone, working with point clouds and modeling the surfaces? I've seen allot about the vertical tolerances and want to know if anybody is using the horizontal stuff for line work, just wondering what the tolerances are between the reconciled photos vs actual field location. If you are interested in pointing me in the right direction and helping me figure out it this is a market I want to enter please email me at chris@newlinesnj.com. I would appreciate any and all input!

Chris Bouffard, post: 420343, member: 12313 wrote: ...costs but can anybody inform me about what the training costs are for operation of the drone,

Using a DJI PHANTOM 3 or 4, you don't need to pay for training. I use map pilot for dji, it's too easy to learn.

I found that using it for linework on a 50 acres site that it worked out really well. My RTK shots vs where I picked from the orthophoto was only .15' different on average and I wasn't being all that careful with my pixel selection when drawing my fence lines.

Sent from my SAMSUNG-SM-N920A using Tapatalk

How did you remove the trees and vegetation from you final results?

geopro_consultants, post: 420387, member: 9959 wrote: How did you remove the trees and vegetation from you final results?

I let AgiSoft do it's initial classification (which was rough, but you may be able to play with the settings to make it better). Then I classified the rest by hand.

Pix4D will create a DSM (Digital Surface Model) and a DTM (Digital Terrain Model), and contours. With each software update (which is about every 45-60 days) the classification and contours are getting better. The user can change classification within Pix4D. You should remove noise such as vehicles, if the software hasn't done this already. When creating the contours Pix4d will also create a reduced point cloud, that is user defined. If I can remember correctly, the general rule of thumb is 1m GSD for 1ft contours, 2m for 2ft contours.

[USER=6939]@Bow Tie Surveyor[/USER]

What type of tablet are you using? Everything I've read suggested the ipad mini 4. But it doesn't seem to have a GPS chip unless you buy one with cellular capabilities?
I doubt I'll ever use "follow me" but would like GPS.

Recommends from Android lovers (me) are more than welcome.

[USER=6939]@Bow Tie Surveyor[/USER] with results like this, I don't see how surveyors focusing on topographic survey (for design or for land title surveys) as a primary part of their business will be able to avoid this and remain competitive. I thought we might have 10 years to transition, but it looks like it is upon us now.

What I see with other existing technology (primarily RTK) is a wide disparity in application. Many still try to operate by rules from 5-10 years ago and those rules may not apply anymore. Photogrammetry will probably be the same. Users will either try to do too much with the new technology and get burned on quality or will not push it enough and will get burned on efficiency. It seems obvious that photogrammetry will not replace field inspection. There are too many things that must be investigated on the ground to really figure them out. But the photogrammetry seems like it will replace the bulk data collection that requires so much time on the ground.

I believe that successful (efficient, accurate, profitable) surveyors will find the right balance between what needs to be field gathered (total station, rtk, level, tape, etc.) and what can be located by aerial.

You should check out lp360 for point classification. It has a lot of aoutomated task. I think they have a uas edition that limits the size in acres of point clouds.

Very cool! Good job and thanks for sharing.

Shawn Billings, post: 420640, member: 6521 wrote: with results like this, I don't see how surveyors focusing on topographic survey (for design or for land title surveys) as a primary part of their business will be able to avoid this and remain competitive.

In my area, +/-0.1' in the vertical isn't good enough for hard-surface topo -- it's just too flat, so site drainage and gravity utility design depend upon more accurate results. (Typical gutter slope is 0.003, same with sewer, and drain is even flatter.) It's the same reason we don't use RTK for topo. I don't see the total station disappearing anytime soon around here.

[USER=10]@Jim Frame[/USER] - no argument from me regarding that. Clearly a case of picking the right tool for the job and in our business it is impossible to ascribe absolutes to how we carry out our responsibilities (which is why I enjoy field work so much). I agree that with sort of slopes, I wouldn't use RTK either, nor aerial photogrammetry.

I think these drone techniques will be "the way" that commercial ALTA/NSPS surveys will be done. Boundaries will still be done by on the ground surveys. Tracing out buried utilities, too. Paving, overhead utilities, fences, structures (maybe), parking stripes, signs will be done by air though. Topo for development requiring 1-2 foot contour intervals will also probably all go this way if the surveyor can convince the developer to clear the land before performing the topo.

I'm thinking I'd better get started on pursuing this or get left behind.

The sad thing is that this technology (in general) will replace all of the easy jobs. The hard fought jobs under canopy, in the briars will probably still require field work. Meanwhile the jobs you wished could last a little longer, with the mowed lawns and edged paving, will be the ones you wrap up the quickest.