Support Website Contact Support Blog

Orthomosaic not matching Google Earth

Good afternoon to all!

I am fairly new to this and am just starting to get used to the Pix4D software. I am using this tool to try and simulate a pond filling up with water in order to estimate its volume. We flew a pond and had the images processed (p4d). In the Pix4D Desktop application, I was using the “Custom Altitude” option in the Volume Settings and increasing the base surface until the water level rose to the spillway (top left of pictures). I pulled a graphic from Google Earth to see how close the simulation was to the actual imagery. It was off by more than I am comfortable with and am not sure what’s going on. I had expected the water to reach the trees on both sides of the pond at the same time (as shown in GE photo).

Again, I am new to this, so I am wondering if adding some GCPs may have helped (I wouldn’t think it would matter too much because I am only interested in the local data for calculation purposes (I couldn’t care less if this pond was in Australia or Arizona). Does the photogrammetry software have a degree of error that would cause this inconsistency? This pond was flown late in the day (shadows), would that make a difference in the processing?

Granted, these images were not flown on the same day, in fact they are a couple years apart. I would not think the sediment would cause such a drastic change in the fill pattern, but who knows?

Any help would be greatly appreciated. Pix4D seems to be a great tool and I am hoping it will work for the projects I am currently involved in. Thanks!

Hi Daniel.  

You have multiple opportunities for inaccurate data here, and kudos to you for your skepticism regarding your results. 

1.  You’re correct that some needs can be met without GCPs and applying scaling constraints instead, but this isn’t one of them. Your water level simulation and it’s gravity model mean that you need some kind of control because you’re not working in isolation. Even if you have an accurately scaled model, relatively speaking, the absolute elevation values that the water level model is working with could be biased/tilted in one direction. 

  1. Without ground control and just relying on the aircraft’s onboard GPS for geotagging, I’d expect the orthomosaics to be +/-30’ when comparing to google earth.  Larger datasets with more images are often +/-7’.    Just personal observations, here. 

  2. Your elevation values can vary greatly, especially if you’re using DJI drones where the image elevations are pulled from the onboard barometer, not the GPS altitude. 


After re-reading your post a couple times, I think it’s safe to say that your head is in the right place and you’re coming to the right conclusions.  This is a job that needs high quality GCPs if you want to simulate the water level, but you could (arguably) still get an accurate fill capacity from the point cloud. 


if it was life or death and GCPs absolutely weren’t an option, this is what I’d do - 

  1. Lay out scaling targets, including a couple that can be used for vertical scaling. Use something with a bubble level for the vertical targets. Bonus points if you also set some hubs or rebar that could be surveyed at a later date if needed. 

  2. Fly Slow(<10mph) with high overlap in a crosshatch pattern collecting obliques at something like 200’ AGL, then do a second flight with nadir imagery at 225’ AGL, rotated 45 degrees from your previous flight. 

  3. There’s a lot of processing settings that can improve your results, but they wouldn’t be addressing the only problem you’re having. 

Scaling targets, including a couple vertical targets with a bubble level, is about the only way I can think of to have some kind of confidence in what you’re making without GCPs, but I’ve never done it like that.  I’d still want to verify with survey a a handfull of times before sending anything out the door done this way.