Mark, I’ll try anything once! That’ll be tonight - stay tuned…
I had a reply from DD on the same issue. They believe that i should be operating at a lower Ground Sampling Distance. Currently for the Pit I work in, I take off from the highest point for fly overs at 100m, giving 5cm/px. I will try this but I’ve lost two drones this years from colliding with trees on the extremities of the pit. From my point of view I’m having to weigh up losing drones or losing data, may just have to stick to making orthomaps and leave volumes for the pros, or invest in the right equipment and training.
What altitude do they recommend?
Today’s a new day with yesterday’s issues. Last night 194 of 225 pics uploaded, but still not enough to calculate volumes. First attempt was 175 of 225, last night 194. Why? The struggle continues…
@Raven_Lens, my flights are now at 90m (the highest permitted) in order to clear trees and the hills around the pit, and to maintain good separation to the top of the stockpiles. In the past, in other areas, I’ve flown as low as 75m with good results.
Did you upload directly from the SD card or transfer them off the card first, then upload?
I tried something different and transferred the pics from the SD to hard drive, and uploaded to DD from the hard drive. Never done that before - it seems a redundant step and takes up 1.62 GB…!!
Seems like it is a DD problem. We flew a site the other day and uploaded around 200 pics. One right in the middle of the project area was left out. Don’t know why. We reloaded the pics and all were used.
Wanted to cover some of the points above in detail:
Uploading from SD card
We always recommend copying over first, as the SD card is generally slow for reads in comparison. Particularly just before uploading when we are reading all the EXIF data of the images and generating thumbnails. SD direct uploads are more likely to become inaccessible halfway through an upload due to power saving strategies employed by the laptop.
Only some images used in processing:
Until 22 hours ago, we had a pre-processing step that removed any image that was very close to the same location but rotated. This step has now been removed as we didn’t see any statistically significant improvement in processing reliability by taking this action, and the photogrammetry part of our pipeline is capable of making these decisions more intelligently.
So previously, even if you uploaded 220 images, we might have discarded 30 prior to photogrammetry as we assumed they were almost duplicated and would delay processing.
However, the map you processed this morning should not have been affected by this issue. We seem to have 194 images uploaded, of which we able to stitch 174. @northarrow - Did you see images had “failed” on the upload interface, or did it say all images uploaded successfully?
Processing problems in the north of the site
The map processing seems to be taking a consistent 3hrs:20mins for this image set, though it is struggling in the northern section because there is very low image overlap from the images that made it through upload and photogrammetry. Along the northern edge, most locations only have 1 or two overlapping images, of the 4-5 needed to create a really decent stitch. I’ve looked at your flight logs and I can’t see any issues there, so I can only assume that the images taken above the hills and trees were dropped because those dead trees are incredibly hard to stitch and would have been 100ft closer to the camera.
Re-uploading images to create another map
We agree that this situation is frustrating. However, in general we have found that reprocessing the same image set will not get you wildly different results unless the overlap % is very low.
That said, we want to add a feature to let you add additional images to an existing map and reprocess it without having to re-upload the original set.
Hi again @northarrow - I’ve looked at more of our metrics and it looks like you completed three uploads this week one with 238 photos (10th) and two with 255 (11th and 12th). I’m going to spend more time tomorrow digging into what’s going on here. We should be using all 255 now for photogrammetry.
It’s also possible this is causing the stitiching problem and lack of overlap in the northern section:
Very good information! I typically try to do 70% sidelap and 80% frontlap. Seems to be a good compromise between good stitching and an efficient flight.
Should I increase the sidelap for volume measuring missions?
Just be like me homie skip DroneDeploy stop using them even though i paid for pro a year up front but wont be renewing anymore. BUt be like me and stick to Pix4D. I flew 3 pit sites yesterday and whenever i finished 1 site as im driving to the next usually 1 hour away the processing is already done on my laptop. I did three sites fully processed and done. No need to upload etc. But besides that you HAVE MORE CONTROL TO SEE WHATS GOING ON step1 gives you a detailed report , i think drone deploy is crazy to come new to the game and charge over $500 for ground control that you still are not in control of doing while pix4d is $350 for cloud and local processing. Jump ship like me!! You will thank me. Plus volumes has 6 options that it will factor for selecting points. Pix4d has been around since 2011 and you would think competition like drone deploy would charge less. I actually lost 2 customers cuz of drone deploy when the customer started getting marketing emails from drone deploy saying for them to do it themselves and cutt out their middle man. Me. The emails didnt say it like tha but you get the point. So not only did drone deploy cost me more money they inadvertently had two of my customers cutt me out due to their solitcitation just because i made my customer get a username login via their email.
Joey, what area are you working? I have questions; operating in Tampa, FL.
Northarrow - I didn’t see for sure while skimming through the thread, are you using high precision GCPs?
Fully agree with joeyseaseu. We purchased Pix4D two years ago. You have much more control over your work. DD didn’t, at the time (not sure of now), offer the resolution I wanted for GCP’s. We are shooting our ground control down to a thousandth of a foot using RTK and/or total station. DD at the time was only allowing me to input the GCP elevation to a tenth of a foot and the lat/lon coordinates were decimal degrees, I don’t recall the resolution.
I emailed DD about this issue and they said they didn’t support this level of precision. Bottom line, if you are working for a client, I feel much better about the data and volumes I am reporting using Pix4D. They are consistent, repeatable, and verify with a traditional topo survey.
Hello everyone, I sort of have an unrelated question so I understand if it gets ignored or removed, I just thought this is a good group of pilots to ask this to. The question is if land elevation change can ever be too dramatic to map and if so, what is the limits? I have a quarry in mind whose elevation goes from ground level to -450ft in a matter of a few hundred yards with the deepest point on site being -600ft. This surely can’t be mapped…? Or can it?
Hey everybody, I hope everyone had a good Easter weekend, I had a few days off so haven’t been at the computer, and just opened the thread. HOLY! Lots of good stuff here!
@Jamespipe - I’ve been going through your posts and still have a few concerns.
-
As mentioned earlier in the post, this is not the first time flying this pit, and in previous flights, the stitching hasn’t been an issue - even last fall when the trees were bare. Even the March flight provided a result when there was snow cover (see map above), and April, too.
-
Although now procedure as recommended, uploading from the SD card to hard drive, then to DD hasn’t previously been required (I’m doing that now, though, so this point is moot).
-
re-loading the same images returns different maps (is this the pre-processing step you spoke of?)
-
The upload interface displayed all 220 images were successfully uploaded
-
I’m still wondering why the volumes of the piles are increasing. This is where the stockpiles are at the bottom of the pit where sufficient altitude would have been maintained (although 90m for the entire site should be sufficient, as in previous flights)
-
As for GCP’s, I have surveyed a few points around the site last Thursday, and am making some targets this week and will re-fly the pit to test this theory. (Stay tuned) We had a wicked snow storm over the weekend (+/- 34"—yes INCHES!!) so it may be a few days until I get out there - the roads are CRAP!!
@joeyseaseu, I hear what you’re saying, but have had alot of successful flights with DD in the past and don’t want to give up just yet. I’m not being facetious or looking to stir anything up, but am just curious: does Pix4D have a forum like this, too? I’m just wondering why you’re still on the DD forum. Are you using both mapping engines now?
@AWGiles3, from everything I’ve heard, the GCPs are that accurate. @dragonflyAS, can you attest to this?
@DiamondHead - yes it can be mapped, refer to the graphics above supplied by @Jamespipe in regards to stitching and overlap at those varying elevations, and don’t forget the elevations are referenced by your take-off position.
Thanks everyone for chiming in on this.
I see the diagram and get the point of overlap but how can there be adequite overlap if the drone takes off from its ground point and goes to an altitude of 200 feet above its take off point, then travels to the center of the quarry that might be -450ft or more? That means the drone could be as much as 650 feet above ground level when placed above the center of the quarry. 650 feet AGL I would venture to say is much too high to gather any reliable data.
If the majority of the subject area was the lower part of the quarry, I would set the flight altitude low at 60’ (giving allowance of course to any obstacles), set the overlaps at 90-95%, then post-process and eliminate any photos not needed from the top areas before uploading.
Just a thought: Are GCPs even available to me on the Pro Plan? I see they are available on the Business Plan at $75/map.