Lee,
I thought bright days were best as you can lower ISO and ensure good image quality? Do the shadows cause issues in Pix4D?
Thanks, appreciate your posts. Be curious what you got going on in Cincy.
DandaMan, post: 410806, member: 11506 wrote: Lee,
I thought bright days were best as you can lower ISO and ensure good image quality? Do the shadows cause issues in Pix4D?
Thanks, appreciate your posts. Be curious what you got going on in Cincy.
Dan,
Shadows cause problems on all aerial imagery. In Pix4D, I have noticed shadows on a paved road will cause vertical issues. Results in bumpy pavement.
With drone we fly at only 200' to 400'AGL, so we are well below the clouds.
Project with Duke Energy. I can't say anymore than that, they have a lot of rules.
leegreen, post: 410808, member: 2332 wrote: Dan,
Shadows cause problems on all aerial imagery. In Pix4D, I have noticed shadows on a paved road will cause vertical issues. Results in bumpy pavement.
With drone we fly at only 200' to 400'AGL, so we are well below the clouds.
Project with Duke Energy. I can't say anymore than that, they have a lot of rules.
Gotcha. We know how that is. Thanks again.
GMPLS, post: 410786, member: 8404 wrote: I don't think it's set a f2.8 unless you manually set it that way. That's just the maximum for that lens. I believe the all of the P3's and the P4 (excluding the P4P) have the same camera as far as still photos are concerned.
Gregg
From my readings, I understand it this way.
The P3A and P4 have the same 4K camera, with a fixed F-Stop at 2.8
The P4Pro is the first DJI Phantom to have variable F-Stop with a range of 2.8 to 11
leegreen, post: 410901, member: 2332 wrote: From my readings, I understand it this way.
The P3A and P4 have the same 4K camera, with a fixed F-Stop at 2.8
The P4Pro is the first DJI Phantom to have variable F-Stop with a range of 2.8 to 11
Whoops, it appears that you're are correct. That may be why there is no mention of using apeture priority when stitching photos (as I inquired above). Thanks Lee.
Gregg,
It was nice to meet you at the NYSAPLS conference. Sorry it was so short.
Drop me an email. I'd glad to bring the P4Pro over run a few test flights.
leegreen, post: 410914, member: 2332 wrote: Gregg,
It was nice to meet you at the NYSAPLS conference. Sorry it was so short.
Drop me an email. I'd glad to bring the P4Pro over run a few test flights.
Yes, it was good meeting you too. It's always tough to find time to chat between classes at the conference.
I would love to see what that P4P can do. Both in the field and office. I may take you up on that.
Gregg
Finally flew the 45 acre site yesterday. It was sunny, light wind, half inch of snow on ground. Ground control is something we're still figuring out. Previously we've used hard features (conc. joints, manholes) and 18"x18" checkerboards (not very durable in it's present form -- cardboard). We do have some plastic we've used to layout conventional GCP's for aerial firms but have not dusted that off yet. For this go-round (rural, rolling hayfields and cow pasture fields) we used 16" diameter pizza pans painted white with a center painted black circle. Drilled out a hole in the center to align over traverse point or stick a landscape nail through.
So I set out 12 points, around perimeter of site and a few in the middle. Distance between varied from 300' to 500'. Site rolls, about 100' drop to the bottoms. Started on top at 150' AGL. 80% front and 75% side overlap, utilizing the Pilot Map App from DronesMadeEasy. I debated going lower but was concerned with keypoint matching this vary drab and repetitive landscape between images, especially with the small 1:2.3" sensor in the Phantom 3 camera. Flight time was 50 mins. Ended up with 815 or so pictures.
Brought into Pix4D, adjusted for rolling shutter, processed, added ground control. 7 of the points were labeled as GCP and the other 5 were check points. My GCP's have errors < 0.05'. My check points are not great with N,E errors of 0.5' (one of them is off 3.5' horizontal) and elevation errors of +/-1.0'.
This is a recap for my own clarity and hopefully for others to weigh in with their 2 cents. At this point I'm going to continue and produce some contours to check against our field data. Short of re-flying, not sure what else I can do to tighten up the data.
Dan,
I suggest using 36"x36" aerial targets in Iron Cross pattern, and mark it in 5 or 6 images, as shown below.
Look at your Quality Report from Pix4DMapper. Show us the report, it has some important info to help determine problem areas.
Good tip. I like those. Attached is my quality report. I believe this is from my first go-round, before I adjusted for rolling shutter. The ground control was way undersized for my altitude, hard to pick up from the snow on ground, and likely not enough contrast (such as in the iron cross or checkerboard).
I see a few problems.
1) You have NOT marked any 3D GCP's
2) You processed without a template, why? Suggest using the default 3DMap at least for startes.
3) You processed with the keypoint image scale at Rapid, change this to Full.
4) You processed using default camera settings. The DJI P3 has a no mechanical shutter and is prone to rolling shutter error. You should turn on Pix4DMapper linear rolling shutter corrections.
Your report should look like this. Notice the huge difference in keypoints per image. 2607 vs. 65670
Lee,
Looks like I sent the wrong report. I have corrected for rolling shutter, marked 12 GCP's on 3-4 images and then did "Automatic Search" to label other images they appear in.
Irregardless, you have pointed out some issues I need to adjust. Will run overnight with the 3DMap template and report back tomorrow.
DandaMan, post: 411769, member: 11506 wrote: Lee,
Looks like I sent the wrong report. I have corrected for rolling shutter, marked 12 GCP's on 3-4 images and then did "Automatic Search" to label other images they appear in.
Irregardless, you have pointed out some issues I need to adjust. Will run overnight with the 3DMap template and report back tomorrow.
Pix4d instructor informed me Not to use the "automatic search". He recommended marking 3 to 6 photos.
Should not be processing all three steps at same time. Process step 1, check results. Once satisfied go to step 2. Check step 2 results, then go to step 3.
Lee,
Thanks again for the tips. I started from scratch. Marked GCPs in 3-4 images. Total of 12 ground points (9 GCP + 3 CP's). Ran the 3D Maps Template (Step 1 only). My results are much better, see attached. Comments welcome.
I have never had that much vertical error in a project. Probably due to your poor targets.
I'd take a closer look at GCP's 3, 7, 8, 9, and 10. Too much error there, especially the vertical at 7.
Try marking more photos, and un-mark those that make it worse.
leegreen, post: 411751, member: 2332 wrote: I suggest using 36"x36" aerial targets in Iron Cross pattern, and mark it in 5 or 6 images,
The only problem (well two) with 3 foot boards is the size - OK if you can drive to all locations, but hard work if you have to walk there with a bagfull. The second problem is securing against the wind, which probably means another bagload of stuff to carry and they are still vulnerable. I've found that 18" boards show up fine, even when seen from several thousand feet by Google (18 boards around the site for those who have time to spare).
@52.692546,-1.3578552,39m/data=!3m1!1e3!4m5!3m4!1s0x48775873ff2deff5:0xfbbf0269e8fca799!8m2!3d52.6909892!4d-1.3609786"> https://www.google.co.uk/maps/place/Ellistown+Terrace+Rd,+Ellistown,+Coalville/ @52.692546,-1.3578552,39m/data=!3m1!1e3!4m5!3m4!1s0x48775873ff2deff5:0xfbbf0269e8fca799!8m2!3d52.6909892!4d-1.3609786
For long term survival I've laminated two squares of the plastic colugated sheet together, with the corrugations at right angles to each other for strength. 3 inch black tape does fine for a square cross (and also helps to hold the boards together). If vegatatioon growth is likely to be a problem then a large square of black upvc lining membrane pinned down on the area keeps it clear for a few years.
That google photo is so blurry, there is no way to determine the center of target. Actual I'm not even certain if was a target at all.
Carrying and securing the proper targets is just part of job. If you cut corners in the beginning, you may jeopardize the end result. You can also paint the targets on pavement. It the most accurate and professional method that I am aware of. We have been this for many years with photogrametry projects. Plus if give me a visual site of the entire project. I'm also measuring features that the aerial mapping can not see. The helps to verify results. We often return to the site for a final field edit with map in hand.
It is a target - i placed it, fixed to a permanent peg.
I wasn't suggesting that you should put down targets, wait for Google to fly it and then use their photography. The point I was making was that 18" square targets were visible even on Google imagery. We also paint targets on pavement, but you can only do that where there are pavements, obviously. When setting the ground control points we also pick up any covers, etc. which are obscured or put in extra local points to let a total station get in and infill.
You might need 3 foot boards when flying at 10,000 feet - I think the first ones I fixed back in 1967 were 4 foot ones - but not when flying at 400 ft.
chris mills, post: 412060, member: 6244 wrote: You might need 3 foot boards when flying at 10,000 feet - I think the first ones I fixed back in 1967 were 4 foot ones - but not when flying at 400 ft.
There is a difference in seeing a target vs. using a target to mark a GCP. They target you have shown can not be used to mark a GCP, too blurry and it is unrecognizable. You may be able to use an 18"x18" at AGL 400ft. But I'd rather be certain, and not take the risk. I see this a lot. People want to use paper gun targets, or ceiling tiles.
At 400 ft. such targets are quite clear - we typically get residuals of 1-3cm. on the targets we use use as check points: the rest are used as ground control, which is quite good enough for volume measurements - probably 10 were ground control, 8 were check points..
As I said in my last post, the Google imagery was just to show that a board of that size could be seen, not that it was to be used from Google. (Although, since thread a few days ago was asking how close to scale Google was, then the board images would be good enough to check distances to better than a metre if they appear on the imagery).