Is Phantom 4 RTK a Joke?

[consultant]

So I watched 4 videos and read a couple how to's and looked at the User Manuals regarding the Phantom 4 RTK and the associated base station. I'm sort of having a hard time believing what I'm seeing/reading....

Nothing I've seen or read shows the ability to place the base station on a control point with known coordinates and enter those coordinates into the base station (or remote controller). This means, while the P4 RTK will have phenomenal position accuracy relative to the base station, there will still be a high degree of error for the absolute world position coordinates. Everything I've seen and read proposes two solutions:

1. Use ground control points, which is laughable as to me the purpose of investing in an RTK drone is to eliminate the need to do that except to do an accuracy check without including the control points in the processing.

2. Connect the remove to a "Custom RTK Network" (term used in the Ground Station interface) which means you don't even use the base station!

3. Use PPK in which case you don't use the base station and you turn the RTK function OFF!

Of the three above solutions, #2 is the "cleanest" as #1 you don't need an RTK drone for and #3, from what I've read the PPK workflow is pretty involved.

Does anyone know regarding #2 what the process is to find appropriate server credentials to input in (flying in the United States?)

 

[R Martin]

So I watched 4 videos and read a couple how to's and looked at the User Manuals regarding the Phantom 4 RTK and the associated base station. I'm sort of having a hard time believing what I'm seeing/reading....

Nothing I've seen or read shows the ability to place the base station on a control point with known coordinates and enter those coordinates into the base station (or remote controller). This means, while the P4 RTK will have phenomenal position accuracy relative to the base station, there will still be a high degree of error for the absolute world position coordinates. Everything I've seen and read proposes two solutions:

1. Use ground control points, which is laughable as to me the purpose of investing in an RTK drone is to eliminate the need to do that except to do an accuracy check without including the control points in the processing.

2. Connect the remove to a "Custom RTK Network" (term used in the Ground Station interface) which means you don't even use the base station!

3. Use PPK in which case you don't use the base station and you turn the RTK function OFF!

Of the three above solutions, #2 is the "cleanest" as #1 you don't need an RTK drone for and #3, from what I've read the PPK workflow is pretty involved.

Does anyone know regarding #2 what the process is to find appropriate server credentials to input in (flying in the United States?)

I never used a Phantom 4 with or without RTK. I flew an Inspire when I used DJI products. I always set control points and still do, even with an RTK solution. I consider it a backstop in case the RTK network I connect to drops the connection. I currently fly both ways; either connected to an RTK network via a data puck or I set up my base station and it talks directly with the aircraft and corrects on the fly.
The most frequent method is just connecting to the network via the puck. It is rare to have a surveyed point that I can occupy on a job and the RTK network gives me a lot more freedom to set up in the most advantageous spot for that particular flight. Coupled with survey-grade GCPs, you really can't tell the difference between the two and setup is a lot easier, albeit the additional time needed to set the GCPs up.

As for GCPs, they are a safety net and a time saver (yes they are a pain in the butt to set up and take down). But, what is more of a pain? Doing the job right once or multiple flights and/or trips? At least with GCPs set up I know when something is wrong and can correct the error. Without them, you just wasted all the time you spent. Technology fails when you need it most. It makes good business sense to have a backup to fall back on.

 

[Earthman]

GCPs are needed for correcting optical and other unavoidable distortions in the aerial photos, the point cloud derived from the photos, and the ortho photo. This has nothing to do with GPS or RTK. You should always set and use GCPs unless you are willing to live with the photographic and other distortions/inaccuracies mentioned.

Surveyed GCPs can also be used for referencing aerial data to a standard datum/reference frame. RTK can replace GCPs in this respect only.

The base station in an RTK pair does not need to be setup on a known position because it measures its actual position using an average of many readings taken over time.

Once the base station knows where it is, it compares its known position against its measured position, which is relatively inaccurate, and calculates the difference. The base station, which isn’t moving, saves these corrections vs time for post processing/correcting the rovers data, and/or sends them to the rover‘s (drone’s) RTK receiver via VHF or UHF radio link so the rover RTK unit can correct its measured positions in real-time.

The RTK pair of GPS receivers need to be in the same general area so that they are subject to the same errors. The base station can be a portable unit or a nearby unit that is part of someone’s network of base stations. Networked RTK base stations are usually accessed via cell phone-based WIFI hotspot.

This is how RTK improves the accuracy of the recorded position of the drone/photos.

If your data gathering relies on photogrammetry to create the point cloud, you will need to set and use GCPs to get the best accuracy.

 

[dronie326]

So I watched 4 videos and read a couple how to's and looked at the User Manuals regarding the Phantom 4 RTK and the associated base station. I'm sort of having a hard time believing what I'm seeing/reading....

Nothing I've seen or read shows the ability to place the base station on a control point with known coordinates and enter those coordinates into the base station (or remote controller). This means, while the P4 RTK will have phenomenal position accuracy relative to the base station, there will still be a high degree of error for the absolute world position coordinates. Everything I've seen and read proposes two solutions:

1. Use ground control points, which is laughable as to me the purpose of investing in an RTK drone is to eliminate the need to do that except to do an accuracy check without including the control points in the processing.

2. Connect the remove to a "Custom RTK Network" (term used in the Ground Station interface) which means you don't even use the base station!

3. Use PPK in which case you don't use the base station and you turn the RTK function OFF!

Of the three above solutions, #2 is the "cleanest" as #1 you don't need an RTK drone for and #3, from what I've read the PPK workflow is pretty involved.

Does anyone know regarding #2 what the process is to find appropriate server credentials to input in (flying in the United States?)

I think you're spot on. The P4RTK is not near was it was advertised as. Not the cheapest option but if you can afford it, the most streamline solution in my opinion is using the Propeller PPK method. That's what we use and it's been working well. Let me know if you need any more info.

 

[UAV_Mapper]

This has nothing to do with GPS or RTK. You should always set and use GCPs unless you are willing to live with the photographic and other distortions/inaccuracies mentioned.

A few things, for those curious about how RTK/PPK drones can work. First off, I haven't used the P4RTK, mostly because of the issue from OP. Second, RTK/PPK can be used without GCPs and can give you amazing results (<0.2' RMSE) all by itself. I've worked with AirGon's Loki and the eBee RTK drones and we were regularly within those constraints without GCPs. Our accuracies were determined by checkshots, which were placed throughout the project site. The checkshots would have stood in as GCPs if needed, but they weren't. We eventually were able to reduce the number of checkshots/GCPs to a much smaller amount and in a less desirable configuration than if they were just being used as GCPs.

With the aforementioned systems, having a calibrated camera was extremely important. This can be done in a lab, or by using a dense target network in a test field. Those camera calibrations can then be used for all of your processing with that camera (provided there are no hard landings, crashes, or anything else that might alter their parameters).

@Earthman states that the base station will get a known location after sitting over a point for a while. Yes, it will get a point, but will it get your point? The difference being that we often work in different coordinate systems, vertical datums, and even site specific coordinate systems. If you're not putting in the exact coordinate of the base for your project, then it's all relative, but not absolute accuracy.

In summary, GCPs are always a good idea, checkpoints are a must for RTK/PPK surveys. If you test and trust your equipment and know it's limitations, then you can get away with far fewer than you would use if only using GCPs. There's a huge difference between relative and absolute accuracy with respect to your project coordinate system. Sometimes you need one without the other.

 

[Earthman]

A few things, for those curious about how RTK/PPK drones can work. First off, I haven't used the P4RTK, mostly because of the issue from OP. Second, RTK/PPK can be used without GCPs and can give you amazing results (<0.2' RMSE) all by itself. I've worked with AirGon's Loki and the eBee RTK drones and we were regularly within those constraints without GCPs. Our accuracies were determined by checkshots, which were placed throughout the project site. The checkshots would have stood in as GCPs if needed, but they weren't. We eventually were able to reduce the number of checkshots/GCPs to a much smaller amount and in a less desirable configuration than if they were just being used as GCPs.

With the aforementioned systems, having a calibrated camera was extremely important. This can be done in a lab, or by using a dense target network in a test field. Those camera calibrations can then be used for all of your processing with that camera (provided there are no hard landings, crashes, or anything else that might alter their parameters).

@Earthman states that the base station will get a known location after sitting over a point for a while. Yes, it will get a point, but will it get your point? The difference being that we often work in different coordinate systems, vertical datums, and even site specific coordinate systems. If you're not putting in the exact coordinate of the base for your project, then it's all relative, but not absolute accuracy.

In summary, GCPs are always a good idea, checkpoints are a must for RTK/PPK surveys. If you test and trust your equipment and know it's limitations, then you can get away with far fewer than you would use if only using GCPs. There's a huge difference between relative and absolute accuracy with respect to your project coordinate system. Sometimes you need one without the other.

I agree with your statements about coordinate systems.

You can use the free program Agisoft Lens to calibrate the camera lens.

It has been a while, but I've used the program to calibrate cheap GoPro knockoff camera lenses. The process is easy and it seems to work. I download Agisoft Lens from here:

Agisoft Lens (64-bit) for Windows - Free download and software reviews - CNET Download

Download Agisoft Lens (64-bit) latest version for Windows free. Agisoft Lens (64-bit) latest update: July 23, 2013

download.cnet.com

The following brief description is from the Agisoft Lens manual:

Agisoft Lens is automatic lens calibration software, which uses a PC's LCD screen as a calibration target.
It supports the estimation of the full camera calibration matrix, including non-linear distortion coefficients.
Estimated calibration parameters can be saved in human-readable file format for subsequent use in the
software, where precise camera calibration data is required.

Agisoft Lens estimates the following camera calibration parameters:

• fx, fy - focal length
• cx, cy - principal point coordinates
• K1, K2, K3, P1, P2 - radial distortion coefficients, using Brown's distortion model


As I recall, you take a photo of a target on the PC's LCD screen and then upload the photo to the PC and Agisoft Lens analyzes the photo and reports the above calibration parameters.

 

[Flightpro]

I think it's a positive advancement. Works for engineers but some surveyors are a little too paranoid for this concept to be 100% accepted at this point.

 

[Nomesurveyor]

The DJI base is poorly constructed and capable of autonomous positioning only. It is cheaply constructed and will not survive much rough handling.
(You can buy a good, used GNSS receiver for much less than the $3,600 that DJI wants for theirs)
You will do much better by collecting static data at the site (1hz rate) simultaneously while the drone is flying to add to the PPK solution. Bonus if you used a known point with published values. GCP's should ALWAYS be used.
I've been involved in photogrammetry for over 50 years. A good photogrammetrist will "see" an error in your GCP as small as 0.10'....really.

 

[Dave Pitman]

You can buy a good, used GNSS receiver for much less than the $3,600 that DJI wants for theirs

Or, maybe something new?

Reach RS2+

Reach RS2+ RTK GNSS receiver is designed features built-in LoRa radio, LTE modem, and a survey app for iOS/Android.

emlid.com

Regarding a ppk workflow with a static gnss occupying a non-know point, I would be curious to learn about the proper workflow if you have the time. Please DM if so.

Cheers,
Dave

 

[Midsky8778]

So I watched 4 videos and read a couple how to's and looked at the User Manuals regarding the Phantom 4 RTK and the associated base station. I'm sort of having a hard time believing what I'm seeing/reading....

Nothing I've seen or read shows the ability to place the base station on a control point with known coordinates and enter those coordinates into the base station (or remote controller). This means, while the P4 RTK will have phenomenal position accuracy relative to the base station, there will still be a high degree of error for the absolute world position coordinates. Everything I've seen and read proposes two solutions:

1. Use ground control points, which is laughable as to me the purpose of investing in an RTK drone is to eliminate the need to do that except to do an accuracy check without including the control points in the processing.

2. Connect the remove to a "Custom RTK Network" (term used in the Ground Station interface) which means you don't even use the base station!

3. Use PPK in which case you don't use the base station and you turn the RTK function OFF!

Of the three above solutions, #2 is the "cleanest" as #1 you don't need an RTK drone for and #3, from what I've read the PPK workflow is pretty involved.

Does anyone know regarding #2 what the process is to find appropriate server credentials to input in (flying in the United States?)

In many states you can find state-run RTK networks that are accessible to the public for free or a small fee (much less than private RTK network subscriptions). As long as you have minimum 3G cell service and access to such an RTK network, the workflow is pretty simple and accurate.

I can say that the DJI base station is unimpressive in build quality and I would hate to rely on it as my only reference point. You can always setup with the DJI base and a second, higher end base like Trimble - from what I’ve heard, the sole advantage of the DJI base station is ease of use with an accompanying DJI drone.

Relying on cell service and hotspot hardware (typically your cell phone) does introduce risk into the workflow - it’s important to watch RTK FIX and also monitor the drone/controller connection as interruptions here impact correction data. I’ve had to refly sites when RTK signal was lost during capture for too long or there was controller interference (corrections must reach the drone from the controller to be encoded with the image).