Mission Planning using DJI M600 & Sony A7R (II)

[BobvdMeij]

Dear all,

I understand this may not be directly related to Photoscan by itself, but considering other Photoscan users out there are utilizing a similar rig I hope to find some advice and learn from other's experiences here anyway! My apologies for the long read.

We are currently in the market for a new aircraft comprising of the following:

Aircraft: DJI Matrice 600 Pro
Gimbal: DJI Ronin MX or Zenmuse Z15-A7
Camera: Sony A7R or A7R II
Optional accesoires: DJI's D-RTK GNSS module

The M600 Pro platform appears to be a comparatively affordable, flexible and near to ideal platform for the mapping endeavours we envision.

Considering that precision and accuracy is of imperative importance, we are currently exploring what (mobile and/or desktop) Mission Planning applications out there will work with the above setup. However, and here comes the tricky bit, in this respect the inclusion of an external (yet DJI supported) camera as the Sony A7R series is seemingly complicated. The M600 by itself is supported by multiple popular applications such as DroneDeploy, UGCS, Map Pilot and DJI's own Ground Station Pro (GSP). The inclusion of a non-DJI camera, however, is suggested to be either impossible or impractical at most. Based on reading countless posts on auxilliary forums it seems we are definintely not the only ones dealing with similar questions.

It seems that none of the applications (other than DJI's own GS Pro) are capable of automatically triggering the Sony A7R camera on the preset locations/ waypoints. As such, we could opt to include an external (and costly!) intervalometer on the camera that triggers the camera on a preset interval, or distance (if GPS is included). Alternatively, we could use the Time Lapse application on the A7R to make the camera trigger automatically on a preset interval, provided we have defined the right value a priori based on desired overlap, flight altitude and horizontal aircraft speed. Then again, this requires us to initiate the Time Lapse a few minutes before take-off. This causes a lot of storage space and, more importantly, battery life to go to waste without having flown a single feet. And from what I have read on multiple forums, the A7R's battery life is relatively limited when in time lapse mode; i.e. it is not unlikely that the battery will be emptied before the flight is ended.

To make matters worse, even DJI's own Ground Station Pro application is unable to automatically trigger the Sony A7R, regardless of using DJI's own gimbal systems AND a camera that is official supported and recognized by DJI. Fortunately, rumours suggest that this may be circumcented by including DJI's Remote Start/ Stop (RSS) module on the Ronin MX gimbal that can trigger the camera through an IR LED. This, however, will add more weight to the aircraft and further limit flight time. Besides, the RSS module is stated to trigger relatively slow which may force us to severely lower flight speed (and therefore; area coverage) in instances when a high forward overlap is required.

At last there is the issue of Geotagging of images. Although Photoscan can match images without knowing where they were taken, having some idea would seriously cut on processing time through using Reference Image-Pair Selection during Image Alignment. Unfortunately, the A7R by itself does not accomodate a GPS module. On the other hand, DJI does not support writing of the M600's position data do the image files when the camera is not one of DJI's own. We could, for example, include an external geotagger (such as this one), although that would be a rather costly workaround considering the costs of the aircraft itself. A much more (logical) solution would be to export the M600's flight logs and align the position and image data based on the time stamps stored by each, for example using a freeware solution as Geosetter. If this post on DJI's forums is anything to go by, however, extracting the flight logs from the A3 onboard flight controller is apparently a hassle, not to say impossible. To make matters even more complicated; does anybody know if it would be possible to extract the positional data from the D-RTK GNSS module (provided that we install that on the bird)?

In short then, the seemingly attractive M600 Pro and Sony A7R combo for mapping applications seems to impose some serious challenges after all! We would easily opt for alternative (non-DJI) industrial grade RPA systems (which offer a full-integration of hardware and mission planning software out of the box) out there, if it were not for the fact that these systems come at a price that is easily 3 or 4 times higher compared to the above setup.

Unfortunately our local DJI distributor is not capable of temporarily lending us an aircraft, let alone the camera system, to put it to the tests a priori. Likewise, there's only a handful of M600's operational in our country and none of them is employed for mapping purposes, hence learning from colleagues or competitors is also impossible. I am seriously hoping there are people out there on this forum that are operating a comparable rig for similar purposes, that are well willing to share their experiences and thoughts. Though relatively affordable, the costs are simply still too high to justify purchasing the aircraft without having any clue on whether it will give a run for its money!

Thanks in advance, any help is much appreciated!

[SAV]

Hi BobvdMeij,

I have used larger DJI gear in the past (my setup: S800 EVO, A2 flight controller, Zenmuse Z15, DJI iPad ground station, Sony Nex 7) which had similar 'issues' as you explained in your previous post.
IMO DJI is great as long as you stick to their 'integrated' setup, which means as soon as you add cameras from other brands (such as Sony) it all gets tricky in regard of UAV photogrammetry. For example, I had to attach a third party IR intervalometer to the Zenmuse gimbal that allowed me to trigger the Nex 7 camera at a set interval (which works, but it is definitely not a very elegant solution). Additionally, the photos were not geotagged, hence I had to use ground control points all the time.

Here is what I would do (2 options):

1. Build my own drone based on the Pixhawk flight controller that allows you to customize everything. If you know how to build drones, it can be done quite cheaply because the parts are not very expensive. It will take some time though and your drone probably won't look that pretty

2. This would be my preferred option. Instead of the M600 package, get a DJI M210RTK with X4S or X5S camera. If you stick to 'standard DJI hardware', it will all work out of the box. I know that the Sony A7R camera has a much bigger sensor which means you could fly at higher altitude and still achieve a very good GSD. However, the M600 that would be carrying a Sony A7R has a much smaller flight time. With a DJI M210RTK you would simply fly at a lower altitude, which means you would need more flight lines to cover the same are at a comparable resolution, but that's not a problem because the battery life is much better on this (smaller/lighter) system. Another advantage would be that you would not haul around that much (heavy) gear compared to a M600 setup.

Price for the M210RTK including camera is about US 15,000 - US 20,000 (about AUD 25,000 + AUD 3,500 for the X5S camera). Considering other competitors, I think that the price is OK.

Note: I've not used an M210RTK myself, but it is the system I would go for. At the moment I am still flying a DJI Inspire 1. It's 2.5 years old now, never missed a beat and never crashed (still original props installed). Great/reliable machine.

If you don't need RTK and a full frame sensor, then a (cheaper) DJI M200 or DJI Inspire 2 might be a good alternative too.

Last but not least, just a thought on UAV system size. Switching from my S800 EVO hexacopter setup to an Inspire 1 meant reduced picture quality, but from a practical point of view all UAV operations became so much easier and quicker to set up. Additionally, the whole gear is much cheaper too.

In the end it will depend on what you are trying to achieve with the gear. Please also consider that if you use a Sony A7R or A7RII camera which shoot at 36.4MP and 42MP respectively, you will need to invest in some really gutsy hardware that allows you to process these images at full resolution (looooots of RAM, several GPUs and multicore CPU). In my opinion, a setup that has a camera with lower resolution (such as DJI's X4S or X5S) is the sweetspot if you also consider the processing side.

I hope you find my insights helpful. All the best for your work.

Regards
SAV

[BobvdMeij]

Dear SAV,

Thanks for your extensive reply, you mention some very valid points of interests.

We have been looking at the M200 series as well, including the 210RTK variant. It definitely has a lot to offer in terms of hard- and software integration, flight time and portability. Then again, data quality is absolute key in our field, even if that comes at the cost of portability. Obviously, the top-mount adds another distinctive and plausibly very valuable feature as well!

Unfortunately, however, DJI's own camera rigs will simply not do the trick. As you mentioned, the GSD will become much coarser. This gets even worse when you consider the X5S has an electronic Rolling Shutter, hence we would have to opt for the X4S and fly even further down. In comparison to a Sony A7R with a 35mm lens at 120m AGL, we would have to fly all the way down to 60m (!) AGL to achieve the same GSD. Although this may be partially compensated for by the extended flight time, we are still bound to VLOS and a maximum distance of 500m from the pilot, hence relocating to subsequent start- landing locations is inevitable after all. Likewise, the image footprint of the X4S is also substantially lower at 60m AGL, forcing additional flight lines.

Building a rig of our own is also next to impossible due to legislation. Our CAA requires a thorough (and very costly!) inspection of the craft, including a documentation check, ground inspection and several flight tests by qualified CAA experts. Consequently, none of the rigs currently operational within our country are currently self-build.

As far as geotagging is concerned, we are well aware of that. We will be using Ground Control (and Check) Points all the time anyway, as this is simply imperative to be able to achieve the accuracy and precision our clients demand. Absolutely accurate geotags are not required anyways, all that we need are relative positions between image pairs to speed up processing during image alignment. Thereafter we get rid of the geotags anyway and base our processing solely on GCPs anyway.

At last we are well aware of the processing constraints. Although you should also consider that a 36MP Sony A7R at 120m AGL produces considerably less images images than a 20MP Zenmuse X4S from 60m AGL. We also believe that we will not be processing thousands of images but rather several hundreds of a time. At this point we do not expect to map areas larger than 50ha each, due to 1) legislatory constraints on the one hand and 2) the highly densified country we are operating in. Nonetheless, we have recently invested in a sizable processing rig comprising of a i7 6850K (6C, 3,6Ghz, 15mb Cache) CPU, 64GB of DDR4-2400 RAM (extendable to 128GB) and two GTX 1080 GPUs. Though not top of the bill, we reckon it will do for now. If it still follows short we are willing to upgrade (or add) some components.

I will keep information in mind nonetheless. Like I said, you mentioned some very meaning- and thoughtful issues. It is a true challenge to find a system that aligns with all your needs in terms of user friendliness, (camera) quality, durability and, obviously, costs. We have also considered more industrial grade solutions as the Altura Zenith ATX8, Asctec Falcon 8 and MD4-1000. These systems largely circumvent the issue of integration as they are interoperable solutions right out of the box. However, their costs are easily triple compared to the above M600 Pro set up. Question is whether that massive difference in price is sufficient to justify losing out on some user friendliness and integration.

[SAV]

Dear BobvdMeij,

It is definitely not an easy job to find the right gear. I've been in a similar situation as you are now a few years ago 

You mentioned that data quality is absolute key for your operations (please have a look at the very end of this post before reading on). Therefore, you might have no other choice than the M600 Pro if you don't want to exceed/blow your budget.

However, have you considered the Canon 5D Mark IV - Ronin MX - M600 Pro combo instead of the A7 - Zenmuse - M600 Pro package? I admit that this will lead to lower resolution photographs (30.6MP images), but according to DJI you should get geotagged images right out of the camera. The latest release note for the DJI M600 Pro states following:
Added GPS information recording when taking stills with a third-party camera connected to a Ronin-MX and Matrice 600 Pro

The Sony A7 does not seem to be supported by the Ronin MX gimbal system. All other supported cameras that are listed on the DJI website are great for cinematography, but not really suitable for UAV photogrammetry. In my opinion, the camera that would work best is the before mentioned Canon 5D Mark IV. The GS Pro app should work too because it allows you to choose a 'custom camera' where you can define all properties (sensor resolution, sensor size, focus length, etc). Not sure if it would allow you to trigger the camera though.

Regarding ground control points. If you are intending to collect them on the ground anyway, then, from a mapping accuracy point of view there is not much value in getting the optional RTK system for your M600 Pro. It will just add weight to the machine and as a result reduce your maximum flight time.

I would not go for one of the more 'industrial grade' solutions that you outlined, and that's not just because of the high price. I think that technology-wise most of the competitors are still trying to catch up with DJI (e.g., military-grade video/signal encryption, obstacle avoidance etc.). And in this quickly developing UAV industry I would not want to invest 100,000s of thousands of dollars for something that is (already) outdated or outdated in 12-24 months (unless you have some really big projects lined up that give you a quick ROI).

You seem to have a quite powerful computer to generate spatial datasets from your aerial images. However, you might still struggle to process several hundred 42MP in Photoscan if you generate dense point clouds at maximum quality settings. But you could always add another GPU later if that's the case. You might also want to get more RAM to not have to split your project up into several chunks for processing.

Coming back to the M600 Pro you want to purchase. The M600 Pro, RONIN MX and Canon 5D Mark 4 is probably the best quality system with proper hard- and software integration you can get from DJI at the moment.

BUT

If image quality (which is not necessarily related to data quality; see very end of this post) is not your #1 priority, I would go for a DJI M200/M210RTK with an X4S camera. Smaller, cheaper and more user-friendly package compared to the M600 Pro setup which still allows you to acquire good quality aerial photographs. Note that the X4S camera uses the same Sony 1 inch sensor as the Phantom 4 Pro, as far as I know. The P4Pro does not look very professional, but IMO it is the best value for money at the moment. I have replaced my old and trustworthy Inspire 1 for a Phantom 4 Pro. Why?

• Much better camera (20MP, 1 inch sensor, mechanical shutter)
• Great flight time (>20 min under real world/windy conditions)
• Everything (including 5 batteries) fits in a single/small backpack
• Batteries can be charged in a car (optional DJI 12V car charger)
• Very affordable system
• Easy to travel with by plane (carry-on baggage)

The resulting photogrammetric models processed from P4P imagery look great and are highly accurate (I generally include 10-20 GCPs that are measured using a survey grade RTK GPS with cm-level accuracy).

Even though you are looking for highest data quality, I would still suggest to trial a smaller/cheaper system such as the M200/M210RTK + X4S or Inspire 2 + X4S. It should be no problem to survey an area of 50ha (1000m x 500m) with a smaller camera/UAV at lower AGL/flight height and still achieve a resolution that matches the one you would get from a M600 Pro setup. Yes, you will have to fly additional flight lines and you'll collect more images. But 1) battery life is better for a small UAV so you easily fly more lines and 2) given that you want a certain GSD (= pixels per cm), you are always collecting a similar 'amount of total pixels', only the number of images changes (less pictures if they have higher resolution). Therefore, it won't be a real difference from a processing time point of view.

From my own experience, I would opt for the user-friendly/fully integrated package. In the past I was always going for highest quality hardware. It seems that I've changed based on my positive experience with the Inspire 1 and P4Pro ;-)

To sum it up. Have a look at the M600Pro - Ronin MX - Canon 5D Mark IV package if you want to get maximum image quality and system integration. You'll have to see if you can trigger the camera from the GS Pro app though. However, if you want a simple/clean and portable system that is supported by (great) third party apps (FlyLitchi and MapPilot are my personal favourites), get a M200/M210RTK+X4S, Inspire 2+X4S or even a P4Pro.
M200/M210RTK is probably the best industrial grade solution because it allows you to add various DJI sensors for other applications such as inspection (Zenmuse Z30) or TIR mapping projects (Zenmuse XT). The M200/M210RTK also has a great maximum flight time of over 30min and obstacle avoidance built-in; the M600 Pro won't give you that.

Coming back to what I said at the beginning of this post, it is definitely not an easy job to find the right gear. I just hope that I was able to help and did not add more confusion ;-)

Regards
SAV


P.S. Just a note on 'data quality' when it comes to UAV photogrammetry. Not only the hardware (camera + lens + drone) but also the following points will have a major impact on the quality of your modelling results.

• Survey layout/design (e.g., image side and forward overlap, use of tie-lines, camera orientation, etc)
• Camera settings (e.g., aperture, ISO, shutter speed, white balance)
• Photoscan processing parameters/workflow (e.g., sparse point cloud optimisation, number of GCPs/check points, etc)

I've seen people using high-end hardware but still getting inaccurate/low-quality modelling results because they don't understand what digital photogrammetry/computer vision algorithms are doing with the photographs to generate 3D point clouds and associated spatial datasets. It is worth to invest some time/money into the workflow side of UAV photogrammetry to guarantee reliable and accurate results.
I have over 10 years of experience in this field and I have done many UAV surveys in various environments; I am happy to assist/consult. Just PM me if you want to talk/chat about it.

[BobvdMeij]

Hey there Sav,

Once more, thanks for taking the time to share your knowledge and opinion on this matter! I do full agree with most of the things that you stated and can get into every argument you make, cheers!

I was somewhat struck by your statement stating the Ronin MX would not support the A7 though. I have been scouting the web in the past few days to find information on this issue and came across multiple rigs that integrated the A7(R) with their Ronin MX. In order to trigger third-party camera's DJI has released a Remote Start/ Stop (RSS) module, which allows one to use the trigger/recording button on their RC to trigger/activate third party camera's. Have a look at its user manual and you will find that with respect to Sony camera's, the Alpha 7 series actually IS supported by the RSS module (as are the Canon 5D/6D/7D). As the RSS only fits on the Ronin MX, that makes the A7 similarly compatible with the Ronin. You can find a demonstration of its integrated workings here. It is worth mentioning that 90% of all industrially produced UAV systems for surveying mappings include an A7R by default, or it is suggested as thé default camera system to include for such applications. That must say something as well...

Regarding RTK we are on the exact same level. We have done research on the matter I found the system is not an actual RTK system, but rather a Differential GPS system. The base station logs its position continuously but is not able to receive or process real-time correction parameters. This will allow cm-level positioning of the M600 in a relative sense to the base station, but NOT in an absolute sense to real world coordinates. It will definitely help during inspection projects though due to reduced magnetic interference.

With respect to the portability and impressive quality of the P4P I could not agree more either. We are already operating an Inspire 1 and X5 camera for survey purposes. But that was only because the P4P had not been released, which it had been two weeks after we purchased our Inspire. Given its incredible sensor (almost similar GSD at the same height), double the flight time and lower weight, it definitely looks like the better alternative of the two! Then again, you will obviously lose out in lower light conditions due to the smaller sensor size and pixel pitch.

One thing I have not mentioned, but which I should have, with regard to our interests in the M600 platform is its payload capacity and relative ease of expandability. We are seriously considering to include a heavy LiDAR system in the coming two years, at this point the M600 represents one of the only platforms that is reasonably affordable and capable to carry such a heavy payload. Likewise, it theoretically allows all sorts of dual- or triple camera setsups, although that will take some creativity to make it happen. I'll give you that.

At last you mention some very relevant points regarding data quality, fortunately we are well aware of all of these. We are a specialized geodetic company with several decades of experience in airborne photogrammetry and remote sensing, as well as in geodetic principles, equipment, workflows, and software packages. Consequently, we are purposefully researching how various parameters in the software operate and how changing each translates into different quality outputs. We are about to conclude a study where we have tested an array of different GCP densities/ layout patterns, as well as the inclusion/ exclusion of transverse overlaps perpendicular to the along overlap at the outer ends of each photogrammetric block, as well as the use of RTK baselines that have been logged for several hours, to see how they may influence/ affect output quality. Later on we will devise similar studies on the influence of different front- and sideward overlaps, as well as flying altitude and camera settings to narrow our workflow down.

Flying the bird itself is relatively straightforward, anyone can do that admittingly. The challenge is to devise and deliver a logical, durable and feasible workflow that efficiently delivers the quality that each client demands.

Cheers!

[SAV]

Hi again BobvdMeij,

Sorry for the confusion about the Ronin MX and Sony A7. I got my information from the DJI website, where DJI only lists the Red Dragon, Arri Alexa Mini, BMCC, Canon 5D, Hasselblad X1D and Hasseblad H6D-100C as compatible cameras, but not the Sony A7.
Here is the link: http://www.dji.com/ronin-mx
Interestingly, in the DJI manual for the Ronin MX the list of compatible devices is much longer and includes the Sony A7 series 

Regarding LIDAR. For a professional grade LIDAR you'll be paying A LOT of money (> EUR 100,000). The market leader RIEGL offers the mini VUX-1 UAV or VUX1-UAV sensors.
Unless you are OK with a non-professional system, e.g., from Velodyne. But it looks to me that you are aiming for best data quality, hence Velodyne is probably not really an option.

If you don't mind, please share what your final setup will look like here in the forum (and how it performs).

All the best.

Regards
SAV

[BobvdMeij]

Hi again BobvdMeij,

Sorry for the confusion about the Ronin MX and Sony A7. I got my information from the DJI website, where DJI only lists the Red Dragon, Arri Alexa Mini, BMCC, Canon 5D, Hasselblad X1D and Hasseblad H6D-100C as compatible cameras, but not the Sony A7.
Here is the link: http://www.dji.com/ronin-mx
Interestingly, in the DJI manual for the Ronin MX the list of compatible devices is much longer and includes the Sony A7 series 

Regarding LIDAR. For a professional grade LIDAR you'll be paying A LOT of money (> EUR 100,000). The market leader RIEGL offers the mini VUX-1 UAV or VUX1-UAV sensors.
Unless you are OK with a non-professional system, e.g., from Velodyne. But it looks to me that you are aiming for best data quality, hence Velodyne is probably not really an option.

If you don't mind, please share what your final setup will look like here in the forum (and how it performs).

All the best.

Regards
SAV

I certainly will! Regarding LiDAR, we are indeed aware this will probably set us back between 100-150k. On the other hand, prices will probably drop in the one or two years to go as development is taking off recently.  It will remain a sizable investment nonetheless, hence we will first focus on visual photogrammetry and as demand increases consider if adding LiDAR is worth the effort and $.

On short notice we may give the Velodyne Puck a go. Although small and performance wise in no way comparable to other Velodyne or VUX systems, it may do the job for certain applications and clients. We are also well connected to a research institution within the country that recently acquired the RiCopter, we are planning to set up a joint experiment before the end of the year.

[MZPCS]

What you want to do is possible and being done.  Contact Drones Made Easy.  They sell a triggering package for the M600 for Sony A7R and it works with other Sony mirror less models too.  You can do precise mapping missions with a pro level camera that provides consistent high quality images without the exposure or white balance issues (to name a few) the DJI cameras are plagued with.  It will trigger reliably down to 1 second intervals.  With that speed and large sensor it is possible to cover a lot of area at high resolutions.

[mobilexcopter]

Hello all,

nice reading, but should someone really use Ronin gimbal for photogrammetry and had the need for accessing camera's hotshoe contact for logging camera's triggering event ( with external logger apparently  ), we have a hotshoe adapter that fits the gimbal at mobilexcopter.com.

BobvdMeij : One note about battery for A7R ( or other cameras where battery life is an issue ) - you can use a power adapter ( battery coupler ) instead of battery and use external ( main ) battery to power the camera.

I would also suggest usage of direct triggering, not IR trigger option for photogrammetry usage. especially if you are flying fast.

[cioro]

Hi,

thank you all for your good advice.
In my opinion choosing A7R/A7R II as a payload for multicopter for photogrammetry is not a smart choice for the following reasons:
- weight, body only is 620gr and this payload will need bigger gimbal also
- 5 axes image stabilization, this is no good for processing because each image will have different camera internal parameters and the processing software will approximate them with a greater error than will do for normal camera.
- large sensor is not needed. same GSD can be obtained by flying at same altitude with smaller sensor but having a lens with longer focal distance
- (not sure) A7R, A7R II cannot take one picture per second
- shorter battery life
- expensive payload
I would have a look to A6300 which has a great sensor in it, no image stabilization, body weights only 400gr and can take at least 1 picture per second which means that you can fly lower (to get almost the same resolution as you will get with A7R) but faster in order to get the same frontal overlap.  With this camera carrying a prime lens you can do very precise point clouds with proper image overlapping.

[MisterGosu]

Hi!
Were you able to try it out in the mean?
All the best!

[BobvdMeij]

Hi!
Were you able to try it out in the mean?
All the best!

As a matter of fact, we haven't. Obtaining a RPAS Operator Certificate in the Netherlands embodies a very costly and length process that can easily take up to two years. We are still in the phase of selecting the most suitable aircraft and payload, so we can take all relevant documents and procedures thereof in our Operations Manual. Furthermore, the aircraft will need to partake in a lengthy technical examination, done by external reviewers and experts, before it may receive its Airworthiness certificate and before it is allowed to be operared. That will take some time still, unfortunately.

Either way, we have currently switched our focus to the M210-RTK after all, particularly for asset inspections. For mapping endeavours we plan to utilize the X4S on the M210.

[SAV]

Hi BobvdMeij,

Do you have any detailed information about the M210's RTK system? Is it used for navigational purposes (like it is the case for the M600 Pro RTK) or is it also possible to store cm accurate location information in the images themselves in real world/absolute coordinates (not just relative to the base station)? To do so one would need to be able to get cm accuracy location information for the RTK base station too (i.e., via CORS). Alternatively, one could place the RTK base station on top of a known point and then update the RTK base station with it's coordinates. Is that possible?

Regards
SAV

[BobvdMeij]

Hi BobvdMeij,

Do you have any detailed information about the M210's RTK system? Is it used for navigational purposes (like it is the case for the M600 Pro RTK) or is it also possible to store cm accurate location information in the images themselves in real world/absolute coordinates (not just relative to the base station)? To do so one would need to be able to get cm accuracy location information for the RTK base station too (i.e., via CORS). Alternatively, one could place the RTK base station on top of a known point and then update the RTK base station with it's coordinates. Is that possible?

Regards
SAV

Hello SAV,

We do have some additional intel, but it is relativley slim. You are indeed correct though that it is not actual RTK in the sense that the absolute position's of the aircraft are stored at cm accuracy. The only position that is cm accurate is the relativele position of the bird with respect to the base station. The base station is only capable of logging and therefore minimally correcting data over time, but it is not provided with a means to receive and process correction signals. From what we have heard it is also not possible to use our Leica and Trimble RTK systems to the M210. DJI's D-RTK is a completely closed system, unfortunately.

There might be a slight workaround though, to at least approximate dm level accuracy geotagging. We are planning to measure the location of the base station using our own RTK-GPS systems a priori, then put the base station in place and conduct a final measurement of its locations afterwards once removed. Alternatively we may employ a tachymeter to measure the base station's position as it is in place.

It took a while but we have also finally received some details from DJI how one is capable to extract GPS data from the M600 Pro's flight controller. It is apparently possible to extract a .txt file through DJI Assistant 2 that stores the GPS XYZ coordinates each time image capturing is triggered. If D-RTK is installed on the M600, it supposedly stores the RTK positions in the same .txt file as well. It is stated that this works while using the Ronin MX, when no GPS data is stored in the EXIF file, which suggests this should also work when using (supported) third-party camera's in conjunction with the M600. I'm not sure if ánd how RTK trigger data is stored on the M210, but I reckon this should be automatically transfered to the EXIF file as it is only compatible with DJI's own camera's at this point, which should therefore be a relatively straightforward process compared to the workflow for the Ronin MX above.

Provided one has the (approximated) RTK position at triggering, as well as the absolute position of the base station (through the above mentioned work around), I reckon one should be able to distill the base station's absolute offset. Which may then be used to correct the RTK data of the aircraft to a more absolute position. All that is left then is to define the offset between the aircraft's RTK system on top and the camera's projective centre underneath. That might be impossible due to constant movement of the camera and it's projective centre, but you should be getting pretty close.

[SAV]

Hi BobvdMeij,

Thanks for the insight.
Would have been great if one could simply connect to a survey grade RTK base station such as the Trimble R2 to get absolute coordinates for the camera stations.
Does DJI give you access to the RINEX file of their base station? This would potentially allow you to compute an accurate location measurement for it based on PPK.

Even when you measure the absolute location of the DJI RTK base station, it will still be quite tricky to estimate the exact location of the camera, as you already mentioned. I guess you might be able to get down to 10-30cm, which for some cases should be OK.

Too bad that DJI missed the opportunity and allow communication to third party RTK base stations to make it a truly professional (survey grade) system. Maybe they wanted to keep that for the next Matrice series ;-)
Nevertheless, I think DJI's UAVs are great pieces of technology for a fair price. I have used them for many years now (since they released their first UAV, the Phantom 1) and they have always performed really well.

Regards
SAV