Pix4dcloud and accuracy?

[JDL]

Hey everyone, so I’ve been messing around with pix4dcloud for my business. Now I’m not a surveyor nor do I plan trying sell survey grade services as I cannot and have no desire to do so. With that said, I do plan on utilizing orthomosaics for construction clients, etc. I recently mapped a parking Lot at my church, I put down a tape measure pulled to 20 ft, the length of one of the lines. I was pleasantly suprised to see that using the measuring tool my approximate length was 19.9738 ft. To me that’s close enough to use for approximations. And that’s without using gcps? Any thoughts on this, I welcome opinions.

 

[R Martin]

Hey everyone, so I’ve been messing around with pix4dcloud for my business. Now I’m not a surveyor nor do I plan trying sell survey grade services as I cannot and have no desire to do so. With that said, I do plan on utilizing orthomosaics for construction clients, etc. I recently mapped a parking Lot at my church, I put down a tape measure pulled to 20 ft, the length of one of the lines. I was pleasantly suprised to see that using the measuring tool my approximate length was 19.9738 ft. To me that’s close enough to use for approximations. And that’s without using gcps? Any thoughts on this, I welcome opinions.

Do you want pictures or do you want maps? There is a difference. Maps in construction are generally made with GCPs to tie the orthomosaic to specific spots on the surface to be used for mapping purposes. Pictures don't require that (but may be geotagged which does tie that image to a specific location). Measurements can be taken from either. They would be a little more accurate with GCPs.

 

[JDL]

Thanks for responding, it answers my question that I can take measurements without gcp’s. although not as accurate as using gcp’s, it seems it’s pretty close, enough for clients to get a ‘rough’ idea.

 

[adm_geomatics]

Do you want pictures or do you want maps? There is a difference. Maps in construction are generally made with GCPs to tie the orthomosaic to specific spots on the surface to be used for mapping purposes. Pictures don't require that (but may be geotagged which does tie that image to a specific location). Measurements can be taken from either. They would be a little more accurate with GCPs.

Sounds like maps if he's measuring. We have been using drones in construction a little over 6 years and I will agree that GCP's are a must, but more importantly because the mature end goal is to tie the drone data to a CAD design which is not always the same as a 1:1 transformation from geodetic WGS84 to a cartesian State Plane coordinate system that construction uses. If you are not needing cm accuracy to tie to designs then you only need RGCP's (as we term them) Rough GCP's that are derived from the first map flown. This then makes all subsequent maps near perfectly relative to each other. This is much more accurate than without. The only time we fly without any sort of GCP's is when we are only doing stockpile volumetrics.

 

[JDL]

Sounds like maps if he's measuring. We have been using drones in construction a little over 6 years and I will agree that GCP's are a must, but more importantly because the mature end goal is to tie the drone data to a CAD design which is not always the same as a 1:1 transformation from geodetic WGS84 to a cartesian State Plane coordinate system that construction uses. If you are not needing cm accuracy to tie to designs then you only need RGCP's (as we term them) Rough GCP's that are derived from the first map flown. This then makes all subsequent maps near perfectly relative to each other. This is much more accurate than without. The only time we fly without any sort of GCP's is when we are only doing stockpile volumetrics.

Thanks for responding. I am interested in making maps. What I was trying to communicate was you could get a fairly close to accurate measurement without gcp's. I'm not referring to pin point accuracy of say topography. I'm thinking of the possibility that a client might not need survey grade accuracy but still my find a meauring or area tool to be 'close' enough to get a general idea, that's all. Can you speak more on doing volumetrics without gcps? case uses etc. i'm very interested in how to incorporate this. Thanks for your input.

 

[adm_geomatics]

In my experience once you get to about 10 acres or so you will start seeing some pretty significant warping in the model without Ground Control Points and I don't think you will be very happy with any kind of cut and fill results. I would encourage you to consider a workflow using rough gcps. This would then allow you to do cut and fill progress, but keep in mind that this is only good for Mass cut and fill and does not apply to many construction applications because of the different layers involved in building something like a building pad or roadway. Volumetrics on individual stockpiles really do not need ground control points unless they get very large like over an acre footprint.

 

[adm_geomatics]

Here's a quick post I did on the DroneDeploy forum about the RGCP process. We can now get the required coordinates for horizontal and vertical in DroneDeploy, I don't think Pix4D has annotation tools quite like it so you still might need the QGIS step.

Sub-Meter Map Relativity

There may be times when you want to have your maps align and be able to report progress, but don’t need centimeter absolute accuracy or have GNSS equipment to set GCP’s. This is a process I use for these occasions. You can paint the ground, put down some inexpensive targets or just use existing...

forum.dronedeploy.com

 

[R Martin]

Thanks for responding. I am interested in making maps. What I was trying to communicate was you could get a fairly close to accurate measurement without gcp's. I'm not referring to pin point accuracy of say topography. I'm thinking of the possibility that a client might not need survey grade accuracy but still my find a meauring or area tool to be 'close' enough to get a general idea, that's all. Can you speak more on doing volumetrics without gcps? case uses etc. i'm very interested in how to incorporate this. Thanks for your input.

Think of your final product as an accordion. For really small jobs the distortions are not going to be as great. The larger the job, the larger the expansion and contraction of the bellows and the larger the distortion. GCPs anchor a specific spot on the image to a specific spot on the Earth's surface. Normally you have at least four corners on the perimeter anchored but you also anchor interior points as well to prevent warping. The more points, the more accurate in theory but there are a lot of other factors involved as well. Ground sampling distance (GSD), or the scale of each pixel of the image, is also a factor. GSD is determined by camera resolution and height above ground. If you have a consistent flight altitude in relation to the surface altitude, the better the accuracy. An undulating flight path (in relation to the surface) where your pixel scale fluctuates is really undesirable if at all possible. Most of the prosumer drones are unable to maintain set altitudes and utilize terrain following.
Without all the wind it boils down to this. How accurate does the client want/need and are you capable of delivering it? Some things require higher accuracy than others. On the front end of the job it is up to you to determine what is good enough versus the cost of that specific accuracy and what the client can afford/(and what he really needs). You would be surprised that a lot don't have a clue what they really need.....

 

[R Martin]

Sounds like maps if he's measuring. We have been using drones in construction a little over 6 years and I will agree that GCP's are a must, but more importantly because the mature end goal is to tie the drone data to a CAD design which is not always the same as a 1:1 transformation from geodetic WGS84 to a cartesian State Plane coordinate system that construction uses. If you are not needing cm accuracy to tie to designs then you only need RGCP's (as we term them) Rough GCP's that are derived from the first map flown. This then makes all subsequent maps near perfectly relative to each other. This is much more accurate than without. The only time we fly without any sort of GCP's is when we are only doing stockpile volumetrics.

Right. And again, all of that is client need driven. In a perfect world everything would be survey grade and Walmart priced but ....
For "mapping" where accuracy counts control needs to be solid. For that 'we' (meaning I) use survey-grade control. That is going to be up to the individual operator to determine what they need to accomplish the job, meet client expectations and most importantly get paid (and leave the door open with that client to do it again). Everything else is driven by cost and the end product.

 

[adm_geomatics]

For really small jobs the distortions are not going to be as great. The larger the job, the larger the expansion and contraction of the bellows and the larger the distortion.

Spot on. In kind the lower you fly warping will increase (if you do not increase the number of GCPs significantly) because tie-points are more localized and do not correct errors across the site as well. This is why large maps that are flown at a much higher altitude like 300-400ft are more accurate with 2D measurements the larger they become. The downside is that you lose resolution and the 3D aspect suffers.

Most of the prosumer drones are unable to maintain set altitudes and utilize terrain following.

If you are referring to models like a Phantom 4 Pro or Mavic 2 Pro then there are plenty of ways to utilize terrain following.

How accurate does the client want/need and are you capable of delivering it?

This is what you should always reverse your approach from.

For that 'we' (meaning I) use survey-grade control.

We find that the majority of the time there is no reason not to do this. The only exceptions I can think of are volumetrics of typical stockpiles of material and/or if the client only wants relative change over time, but we should also be prepared for a future request that may require us to go back in time in which survey-grade may have been required. This is why we always at least set some targets even if we don't shoot them. They can be shot later if needed a the previous maps reprocessed.