Significant Volume Differences based on point placement


#1

Hi, I am having an issue where my stockpile volumes change quite significantly depending on the placement of my points (when drawing the circle). For example, I have a stockpile that I have measured with an RTK gps to be 8375 yds. Casually clicking around it in DD (close to the base, and evenly spaced) I got 8227. I deleted that circle and did it again getting 8156. These are relatively small errors, but if I do nothing but just increase the density on one side of the pile while keeping the other “normal” I get 7849 yards. That translates to 6.3% error just in the circling process. Across our whole site that is a lot of inventory value. Is Drone Deploy just taking the average elevation as the base? In which case I need to be very careful to place them evenly around it? Or am I missing something?

I had a colleague of mine fly it with their system and no matter what we did it stayed within 30 yards of the RTK (0.4% error)

I would appreciate if anyone has found “Best Practices” to keep accuracy spot on. I really like Drone deploy and would like to stick with them if I could.


#2

FYI, I use a phantom 4, but I’m not sure that has much to do with the circling accuracy. Never know though, so I figured I would mention it.


#3

Hi @DaleyP,

Thanks for reaching out to us on the Forum! Have you checked out our guide on Accuracy yet? We have useful information about the variables affecting your map accuracy. I’m also curious to know if you added any Ground Control Points to your map? If so, could you please provide the name of your map and the username associated with your DroneDeploy account so I can take a look?

Looking forward to hearing from you!

Best,
Stephanie


#4

Yes I’ve looked at the guide on Accuracy, I didn’t see anything about circling techniques. It appeared to be mostly about capturing the data properly (I may have skimmed it too quickly).

The map in question does not have GCP’s but I have recreated this same phenomenon on my maps that do have GCP’s.

Username: daley.paterson@gcinc.com
Map: REC AB Jan 22
Map with GCP’s if you want to try it out too: Vernalis Plant - All w/GP Oct 18

I noticed the issue a while back, its most apparent on piles that are on a slope.


#5

Thanks for your quick reply! It is common to see more discrepancies in your volume calculations without the addition of Ground Control Points. That is expected given the relative accuracy of your drone’s GPS.

I also think an explanation of how we calculate volume might be helpful here. We add up all the individual pixel volumes to calculate the total volume. In doing so, the uncertainty in your volume estimate is basically a function of how small the pixels are (resolution), and the uncertainty of each pixel height reconstructed by photogrammetry. Additionally, the world isn’t perfectly flat. Since there may be a slight incline to the area in question, we have to subtract out the height of the “base plane” from the elevation value at each pixel. We do this by choosing a 3d plane that best represents the ground at the perimeter (an average), and subtracting that plane from each pixel height. This is likely the reason why you are seeing slight inconsistencies.

I hope this additional context makes sense! In terms of best practices, you definitely do want to be careful about placing the points evenly throughout the stockpile. This should help with accuracy.

Best,
Stephanie


#6

Use the colored elevation map to pick your points. This will ensure the max volume available. Also, you can use the distance tool perpendicular to the toe to find the best grade break and also ensure the low point. This is particularly useful when quantifying multiple piles that are close together.


#7

@DaleyP What was the system that provided relatively tight #s regardless of slight differences in selection polygon?


#8

The drone system or the workflow? I’m not sure I understood the question…:confused:


#9

My question is for Daley. He said his colleagues system had much more consistent results. What system?


#10

@Stephanie Perfect, thanks Stephanie, that helps! I’ll make sure I’m more evenly spaced from now on. I suspected this would be needed but I wanted to make sure.


#11

I have to disagree with results I have seen. In my experience with non-GCP volumetrics while the error is obviously larger than with GCP’s, stockpiles are very isolated and local so it really doesn’t make a huge difference unless they are 20,000+ cy piles. Results I have experienced are within 2-5% of previous flights where the stockpile did not change and were within 5-7% of an RTK GPS survey which had allot more shots than most people would probably take. It was 30 minutes worth of work for the complete drone workflow vs 1.5 hours of foot time and 20 minutes for post processing. That taken into account we have actually tracked trucks on a drone surveyed pile of close to 10K cy and were within 11 trucks (or 130 cy) of what it said. What you cannot rely on is that the pile is going to be in exactly the same place on the next map, but the quantities should be close if you place your basepoints correctly.


#12

I’m also interested in this system. I’ve just gotten into this with drone deploy and am still under the trial period.


#13

There are definitely limitations when using the DD volume tools for stockpile volumetrics, they are really only useful for getting a rough idea of a pile volume, i would say its typically closer to being within 10% of the actual volume in the most ideal situations. Those limitations are based primarily on the fact that you cannot specify the base plane elevation or use the lowest point of your perimeter polygon, both of which would make the calculations far more accurate and useful in most stockpile situations where they are not on a perfectly flat plane or are butted up against a wall or another pile, which is often the case. If you need higher accuracy you will need to download the point cloud and use another program. We have had greater success using something like Quick Terrain Modeler or Pix4D or even Propeller Aero sinc they allow for basic base plane manipulation, but those of course are not cheap options like DD, but of course you get what you pay for. DD is still a great tool for ballpark measurements. Just keep in mind the bigger the pile the smaller the relative error will be, so for small piles it will make a big difference but not so much for big piles.


#14

If the pile is on a flat plane right? We have piles on jobsites so matching the terrain on which the pile sits is great and we find it more accurate that GPS rover surveys according to the trucks…