Support Website Contact Support Blog

GPS Recomendations for GCP


During my first test mission using GCPs yesterday, we used an android phone to record the GPS coordinates of each of the GCPs. When my partner sent the coordinates over to me, I noticed that it only had the X & Y values, not the altitude.

So, I think I either need to get an app for my iphone 6, or buy a GPS. I hope I’m on the right track here.

If I’m correct in my assumption, does anyone have a recommendation for this? If I do need to buy something, preferably it would be an app for iOS.

Also, I’m flying an Inspire 1.

Thanks so much!



Hi Aaron,

Chances are that the precision afforded by your mobile device(s) that are outfitted with consumer-grade GPS won’t be any better than the positions afforded by the Inspire 1. So buying an app for whatever system, iOS or Android, isn’t going to overcome the limitations of this type of receiver-antenna built into your phone.

Depending on your mapping needs, you could even proceed without GCPs reliant solely upon the geotagging done by the DJI Inspire 1 while ever mindful that positions will be approximate; e.g., horizontal uncertainties likely within 3-5 meters and vertical uncertainties varying wildly. We’ve routinely seen ridiculous values up to hundreds of feet from the Inspire 1 between the geotagged altitude and the post-processed results with solidly surveyed GCPs. On other days, and quite unpredictably, differences will be more believable when expectations are tempered by the fact that Inspire 1 uses its barometric sensor to write z values while geotagging.

If you have GIS tools in your kit, access to professionally produced real ortho rectified imagery, good terrain data (LiDAR, DEMs, etc.), you can produce quasi-GCPs from those previously produced spatial products. The quality of that source data will largely impact the end result of your Pix4D project in a similar fashion as flying at different altitudes effects the Ground Sample Distance (GSD).

To achieve the best results, GCPs need to be surveyed. Typically, this will be done using GNSS RTK techniques which requires a rover unit that is receiving corrections from a base unit and its radio transmitter. Ideally, the base is on the project site, but corrections can also be received by many RTK rovers over the Internet using cellular communications and the base station CORS network frequently maintained by state departments of transportation. The local base station being the most robust (and expensive) choice. Alternatively, GCPs could be located using GNSS Static techniques which can produce very accurate positions after its data has been post processed. Static surveying will often mean longer occupation times too but won’t cost as much.

Before buying anything (it’s all quite expensive!) I suggest you determine the level of precision your projects require and then direct energies to learning about the tools needed in achieving those targeted specs. Your greatest dividends will be from the time you invest educating yourself before the purchase.


dont use altitude from a cell phone. Ive tried multiple times as well as using surveyed GCP and importing the image locations under a custom accuracey of 300’. never worked out for me.

 I have spent nearly a week to try and figure out what is the best system to buy to get reasonably accurate GCPs, but it’s a real mind field to figure out what to buy and for what you can pay and what you get for that if your not a professional surveyor in the know.

We basically like to achieve around 1cm-2cm accuracy for the z axis which for volume cut and fill areas is the most critical for the majority of the work we do with Pix4D.  I see the RTK GNSS systems are obviously the most accurate in a short occupation time frame in the field but also the most expensive, where as others will do a reasonably good job with a static base stations, but having to have a longer occupation time frame. Even so the cost can vary from $1000 to $45000 when for the higher end cost it also includes special processing software.  Then the next issue is how some of the systems generate the data that then needs to be converted to use in Pix4D in some systems seem very complex or it needs other external paid services as well.

So where should one start and go from here?

Well firstly it would be good if Pix4D could set out a basic guild line chart of some recommendations similar to how they set out the recommendations for a PC to do the processing.

In the meantime and taking into account our preferred budget would be around $2000-$6000, can anyone please give some type of guidance and recommendations which system would be able to achive the type of this type of acuracy?  

Hello John,

In fact, it would be useful to have an article with some GPS recommendations. We will work on it. 

As a general rule, if you would like to have an absolute accuracy of 1-2 cm for the outputs you should measure the GCPs with survey grade GPS. If not, the the image geolocation and maybe some GCPs extracted from Web Map Services would roughly georeference the model. 

Please note that the relative accuracy of the model (scale) will be high even without GCPs.

Best regards,


How’s the GPS measuring kit recommendations list coming?

This is a real minefield and I think like everyone, we understand we need to invest but we don’t want to make an expensive mistake. This is holding us back from buying a license for Pix4d as we need to be sure we can provide about 2cm accuracy and cost out or projects.

Interested in this subject aswell.

If you have markets that require 2 cm accuracy, then just make sure your market does not include providing data to engineers or architects. That type of mapping generally needs to come from a licensed surveyor or photogrammetrist. I’m not sure what other industries might require that type of accuracy, but I sure hope there are and I’m just not aware of them!

So if your data is just for informational purposes, the GPS from the geotagged images might be good enough. Points on the model will still be very accurate relative to other things close to it in the model, it will still look visually correct. But it might not have the same coordinates for a point on the model as the same spot on the ground, but why should that matter much if it still looks great and it’s just for informational use. This type of error generally builds up slowly over distance, it’s not like the model will have an abrupt visual shift, if it does, you probably did not get enough images.