Please select Into the mobile phone version | Continue to access the computer ver.
SRT Accuracy
244 23 11-4 04:12
Uploading and Loding Picture ...(0/1)
o(^-^)o
cpraschl
lvl.2

Austria
Offline

Hi everyone!

I am currently working in a project, where we use two M30T, a M300 and a Mavic 2 Enterprise Thermal (M2EA).For the project we need the exact drone position, when evaluating the video data.
To do so, we are using the SRT files created by the drones.
Unfortunately, we have recognized that the GPS accuracy within the SRT files varies from drone to drone.

The M300 and M2EA log coordinates with 6 digits, which is fine for our use case.
Compared to that our M30T drones are only logging coordinates with 5 digits, which is unfortunately not enough for our use case.
Is there any possible to configure the digits?

Below you can find exemplarily entries of each drone's SRT file.

M2EA:
1
00:00:00,000 --> 00:00:00,033
<font size="36">FrameCnt: 1, DiffTime: 33ms
2022-04-26 07:31:20,106,409
[color_md : default] [latitude: 48.398149] [longtitude: 14.437132] [rel_alt: 13.818 abs_alt: 577.824] [Drone: Yaw:0.6, Pitch:-1.2, Roll:-10.7] </font>


M30T:
1
00:00:28,628 --> 00:00:28,662
<font size="28">FrameCnt: 859, DiffTime: 34ms
2022-09-12 11:56:31.304
[focal_len: 40.00] [dzoom_ratio: 1.00], [latitude: 48.36793] [longitude: 14.51061] [rel_alt: 23.131 abs_alt: 522.435] [gb_yaw: -85.3 gb_pitch: -99.1 gb_roll: 0.0] </font>


M300:
1
00:00:00,000 --> 00:00:00,033
<font size="28">FrameCnt: 1, DiffTime: 33ms
2022-05-24 08:12:19.056
[iso: 120] [shutter: 1/400.0] [fnum: 1.0] [ev: 0] [focal_len: 58.00] [dzoom: 1.00]
[latitude: 46.782160] [longitude: 13.733278] [rel_alt: 64.232 abs_alt: 888.275]
[drone_speedx: 0.0 drone_speedy: 0.0 drone_speedz: 0.0]
[drone_yaw: 0.1 drone_pitch: 0.0 drone_roll: -0.1]
[gb_yaw: -4.3 gb_pitch: -90.0 gb_roll: 0.0]

0
[ae_meter_md : 0] [dzoom_ratio: 10000, delta:0] [color_md : default] [ct : 5000]
</font>


Best regards,
Christoph
11-4 04:12
Use props
patiam
Captain
Flight distance : 1005991 ft
  • >>>
United States
Online

You speak of accuracy, but what you seem to be concerned about is the number of decimal places in the recorded positions.  There is NOT necessarily a relationship between these two things. Any system can be configured to report as many digits as you want; but the number that are actually meaningful is dependent on the accuracy of the system.

You should be more concerned with whether the aircraft has RTK fixed or not, and the RMS or SD of the position. And even with RTK fixed, the M300 is going to be better than the M30T or M2EA, due to the lack of Time Sync on the latter two.
11-4 19:37
Use props
DJI Natalia
Administrator

Offline

Hi, there.
Thank you for your question, we will forward your feedback to the relevant team and see how we can do better.
Besides, if there is anything else that we can help with, please feel free to let us know.
11-6 23:40
Use props
cpraschl
lvl.2

Austria
Offline

patiam Posted at 11-4 19:37
You speak of accuracy, but what you seem to be concerned about is the number of decimal places in the recorded positions.  There is NOT necessarily a relationship between these two things. Any system can be configured to report as many digits as you want; but the number that are actually meaningful is dependent on the accuracy of the system.

You should be more concerned with whether the aircraft has RTK fixed or not, and the RMS or SD of the position. And even with RTK fixed, the M300 is going to be better than the M30T or M2EA, due to the lack of Time Sync on the latter two.

Dear patiam!

thanks for your reply. First of all, yes you are right RTK is of highest importance for our use case during the flight. But I also need the positional accuracy afterwards when processing the data. And GPS accuracy is directly connected to the number of digits as listed below:

decimal places        degrees        distance
0        1        111 km
1        0.1        11.1 km
2        0.01        1.11 km
3        0.001        111 m
4        0.0001        11.1 m
5        0.00001        1.11 m
6        0.000001        0.111 m
7        0.0000001        1.11 cm
8        0.00000001        1.11 mm
So I need the digits in SRT to accuratley associate a video frame to its global position.

Best regards,
Christoph
11-6 23:46
Use props
cpraschl
lvl.2

Austria
Offline

DJI Natalia Posted at 11-6 23:40
Hi, there.
Thank you for your question, we will forward your feedback to the relevant team and see how we can do better.
Besides, if there is anything else that we can help with, please feel free to let us know.

Thanks a lot Natalia!

Best regards,
Christoph
11-6 23:48
Use props
DJI Natalia
Administrator

Offline

cpraschl Posted at 11-6 23:48
Thanks a lot Natalia!

Best regards,

You are welcome, thank you for your support of DJI as well.
11-7 00:11
Use props
LV_Forestry
First Officer
Flight distance : 4149278 ft
Latvia
Offline

cpraschl Posted at 11-6 23:46
Dear patiam!

thanks for your reply. First of all, yes you are right RTK is of highest importance for our use case during the flight. But I also need the positional accuracy afterwards when processing the data. And GPS accuracy is directly connected to the number of digits as listed below:

I understand what you want do, but i agree with Patiam. More digit after the decimal doesnt mean data are accurate. But you can't convert from WGS to meter simply like that, because 1degrees is maybe equal to 111km at the equator level, but the earth is round.
More seriously, if i have to do this job, i would forgot about coordinates in SRT and concentrate on the timestamp. Then collect DJI log or even better your RTK provider log if it exist. Then connect each frame to the precise and accurate coordinates, according to the timestamp. Otherwise instead of making a video, you put the photo mode in burst at regular intervals. It will be like a video but with less FPS, and especially georeferenced images.
11-7 08:12
Use props
patiam
Captain
Flight distance : 1005991 ft
  • >>>
United States
Online

cpraschl Posted at 11-6 23:46
Dear patiam!

thanks for your reply. First of all, yes you are right RTK is of highest importance for our use case during the flight. But I also need the positional accuracy afterwards when processing the data. And GPS accuracy is directly connected to the number of digits as listed below:

@Christoph-

Thank you for the lesson regarding the relationship of decimal degree digits to linear distance

I assure you that as someone with > 25 years in hydrographic and terrestrial surveying using SONAR, LiDAR, and photogrammetry as well as teaching GIS, I understand this concept. (And I second @LV_Forestry's point that while your numbers are approximately true for both Lat & Lon @ the equator, they become less and less so for Lon as one moves away from it).

But reporting more decimal places in a value does not necessarily mean that they are all significant- any number can be rerpesented with more or fewer digits and adding digits only yields additional precision/accuracy if the system is truly accurate to that level. There are autonomous (non-DGPS) GNSS units out there that will give you your position out to 8 decimal places even thouigh their accuracy is only ± 5 m! In other words, nothing after the 4th decimal place can be trusted, so you're deluding yourself if you think that having more digits means better accuracy.
I hope you understnd this, and consider @LV_Forestry's advice on a more robust way to get your geotags than via SRT.
11-7 10:35
Use props
cpraschl
lvl.2

Austria
Offline

LV_Forestry Posted at 11-7 08:12
I understand what you want do, but i agree with Patiam. More digit after the decimal doesnt mean data are accurate. But you can't convert from WGS to meter simply like that, because 1degrees is maybe equal to 111km at the equator level, but the earth is round.
More seriously, if i have to do this job, i would forgot about coordinates in SRT and concentrate on the timestamp. Then collect DJI log or even better your RTK provider log if it exist. Then connect each frame to the precise and accurate coordinates, according to the timestamp. Otherwise instead of making a video, you put the photo mode in burst at regular intervals. It will be like a video but with less FPS, and especially georeferenced images.

Dear LV_Forestry,

thanks a lot for your input!

We are already also using the DJI log files Our project is a research project with approaches from citizen science integrating hobby pilots. From our experience, accessing the log files is not that intuitive for many hobbyists compared to using SRT files. So, for this reason, we wanted to evaluate the utilization of SRT as an alternative, but without waiving the positional accuracy.
And I am aware that the conversion from radial WGS coordinates to metric coordinates requires some model and is not a simple linear conversion.

Regarding the utilization of photos: Unfortunately, we require the amount of frames for our purposes. So reducing the FPS to some photos does not work out

Best regards,
Christoph
11-11 03:04
Use props
cpraschl
lvl.2

Austria
Offline

Dear patiam,

I really didn't want to deny your expertise. Sorry if that feeling came up.

As already replied to LV_Forestry: I am really aware about all those things, that there are increasing inaccuracies the farther you get from the equator.

And as a computer scientist, I am also aware that more decimal places do not necessarily mean that they are all significant However, since DJI drones are able to fly and land with centimeter precision using RTK, I don't think that the data logged to SRT from the drone itself would be that inaccurate Respectivley, that the relative accuracy from the start position would deviate that much.
Best regards,
Christoph
11-11 03:16
Use props
LV_Forestry
First Officer
Flight distance : 4149278 ft
Latvia
Offline

cpraschl Posted at 11-11 03:16
Dear patiam,

I really didn't want to deny your expertise. Sorry if that feeling came up.
"since DJI drones are able to fly and land with centimeter precision using RTK,"
->Remember that during the flight, the onboard GNSS antenna is in motion.
11-11 04:28
Use props
patiam
Captain
Flight distance : 1005991 ft
  • >>>
United States
Online

cpraschl Posted at 11-11 03:16
Dear patiam,

I really didn't want to deny your expertise. Sorry if that feeling came up.

@Christoph-

Thanks for acknowledging the issue of significant digits. Sorry if I was preaching to the choir there but your initial posts led me to believe there was a need to bring that up.

I'm confused that on the one hand you are speaking of using "citizen science integrating hobby pilots" and the need to make the process more intuitive than log files allow, but on the other hand you state that with RTK, cm precision is possible.

Most hobbyists do not have RTK-enabled aircraft. And those that do can surely understand how to access a log file.

And your original post speaks of using M300, M2EA, and M30T, all of which can be capable of RTK, but are rare to non-existent in the "hobbyist" population.

Once again, without RTK, aircraft positioning is ±2-5 m at best. So nothing after 5 decimal places in a decimal degree coordinate is going to be meaningful.

Even on RTK aircraft such as the M30T and M2EA, the lack of Time Sync means that imagery positioning precision is compromised.
11-11 10:30
Use props
patiam
Captain
Flight distance : 1005991 ft
  • >>>
United States
Online

LV_Forestry Posted at 11-11 04:28
"since DJI drones are able to fly and land with centimeter precision using RTK,"
->Remember that during the flight, the onboard GNSS antenna is in motion.

and its relationship to the center of the CMOS in 3D space is constantly changing...
11-11 10:32
Use props
cpraschl
lvl.2

Austria
Offline

LV_Forestry Posted at 11-11 04:28
"since DJI drones are able to fly and land with centimeter precision using RTK,"
->Remember that during the flight, the onboard GNSS antenna is in motion.

Dear LV_Forestry,

we have that in mind, but thank you Since our algorithms seem to work very well with the on-board log data, this should not be either a big deal for the SRT use case.

Best regards,
Christoph
11-14 02:23
Use props
cpraschl
lvl.2

Austria
Offline

patiam Posted at 11-11 10:30
@Christoph-

Thanks for acknowledging the issue of significant digits. Sorry if I was preaching to the choir there but your initial posts led me to believe there was a need to bring that up.

Dear patiam,

maybe hobbyists was the wrong term. The user group in mind are mostly farmers/hunters that are already using thermal drones for finding baby deer before driving combine harvesters through fields. So they are already familiar with RTK, because it is also needed for their farming vehicles, but they are mostly non-professional drone pilots. And we are already in contact with about a dozen of them and did some usability tests with the finding that getting the log files was a big hurdle for them. This is the reason why we are looking for an alternative using SRT. And this is the reason of the finding with the different number of digits ^^

We have some constraints, like that the camera has to look straight down to the ground and the drone has to be directed to north. So the relationships of the moving CMOS can be simplified to a relationship to the drone position. As mentioned in my previous post, the approach is working fine with log files But we want a simpler alternative to the log files when analyzing the video data.

Best regards,
Christoph

Best regards,
Christoph
11-14 02:32
Use props
patiam
Captain
Flight distance : 1005991 ft
  • >>>
United States
Online

@Christoph-

Ah, makes more sense now. Best of luck with your endeavor!

Cheers!
11-14 07:32
Use props
LV_Forestry
First Officer
Flight distance : 4149278 ft
Latvia
Offline

cpraschl Posted at 11-14 02:32
Dear patiam,

maybe hobbyists was the wrong term. The user group in mind are mostly farmers/hunters that are already using thermal drones for finding baby deer before driving combine harvesters through fields. So they are already familiar with RTK, because it is also needed for their farming vehicles, but they are mostly non-professional drone pilots. And we are already in contact with about a dozen of them and did some usability tests with the finding that getting the log files was a big hurdle for them. This is the reason why we are looking for an alternative using SRT. And this is the reason of the finding with the different number of digits ^^

The approach still seems complicated to me. To detect hot bodies in nature I do it rather with rasters. You take a set of infrared images, alignment, rasterization... You take out a temperature filter that is significantly higher than the average of the raster. And then that's all ! This works for bird nests if they are not too much under the canopy. For a deer in a field I'm not worried.

Because the technique of the video includes that someone has to watch it in its entirety? or is there a software that detects?

If it is an operator who does the detection then the H20T is perfectly designed to do this job. With its telemeter it is used to give the position of a point in space.
11-14 08:29
Use props
cpraschl
lvl.2

Austria
Offline

LV_Forestry Posted at 11-14 08:29
The approach still seems complicated to me. To detect hot bodies in nature I do it rather with rasters. You take a set of infrared images, alignment, rasterization... You take out a temperature filter that is significantly higher than the average of the raster. And then that's all ! This works for bird nests if they are not too much under the canopy. For a deer in a field I'm not worried.

Because the technique of the video includes that someone has to watch it in its entirety? or is there a software that detects?

Dear LV_Forestry,

sorry for my late answer, I have been on vaccation the last two weeks.

Detecting hot bodies in nature is as you said a quite okayish task... As long as they are not covered due to canopy, foliage, etc. And exactly this is the part of our research project. Putting many many geo-referenced video frames together, to get insights through canopy in a forest. Comparable to the work my colleagues already did for rescue operations, but in our use case for animal detection and counting (https://arxiv.org/ftp/arxiv/papers/2105/2105.04328.pdf). So yes the approach is complicated, because our project target is. ^^
11-24 23:52
Use props
LV_Forestry
First Officer
Flight distance : 4149278 ft
Latvia
Offline

cpraschl Posted at 11-24 23:52
Dear LV_Forestry,

sorry for my late answer, I have been on vaccation the last two weeks.

Thanks for this clarification.
You almost need real-time image analysis if I understand correctly.
11-25 08:57
Use props
LV_Forestry
First Officer
Flight distance : 4149278 ft
Latvia
Offline

cpraschl Posted at 11-24 23:52
Dear LV_Forestry,

sorry for my late answer, I have been on vaccation the last two weeks.

There is the same type of project organized here with a company specializing in the automated analysis of aerial and satellite imagery. Feel free to reach them

https://www.edi.lv/projects/sate ... nologija-woodstock/
https://www.forestradar.com/
https://www.baltsat.lv/
11-26 02:50
Use props
cpraschl
lvl.2

Austria
Offline

LV_Forestry Posted at 11-26 02:50
There is the same type of project organized here with a company specializing in the automated analysis of aerial and satellite imagery. Feel free to reach them

https://www.edi.lv/projects/satelitdatos-balstita-jauna-mezaudzes-krajas-novertesanas-tehnologija-woodstock/

Hi LV_Forestry,

thanks a lot for the links I will have a deeper look at it, maybe there are some synergies.
11-27 22:16
Use props
LV_Forestry
First Officer
Flight distance : 4149278 ft
Latvia
Offline

cpraschl Posted at 11-27 22:16
Hi LV_Forestry,

thanks a lot for the links  I will have a deeper look at it, maybe there are some synergies.

You're welcome,

I take part in one of their project only for the supply of data, but from what I can understand yes there are similarities.
11-27 22:55
Use props
cpraschl
lvl.2

Austria
Offline

DJI Natalia Posted at 11-7 00:11
You are welcome, thank you for your support of DJI as well.

Dear Natalia,

I wanted to inform you that I have the same behavior with our newest Mavic 3 Thermal.
So maybe it is associated with the DJI Pilot App V2?

I have to recheck it with the updated M300, because the sample frame that I have posted in my initial post was created with the old version of the DJI Pilot App.

11-29 22:28
Use props
DJI Natalia
Administrator

Offline

cpraschl Posted at 11-29 22:28
Dear Natalia,

I wanted to inform you that I have the same behavior with our newest Mavic 3 Thermal.

Thanks for your feedback and sorry for the inconvenience caused.
We are confirming this problem with the relevant team, once we get any updates, we will reply to you here soon.
Thank you for your patience in advance.
yesterday 00:42
Use props
Advanced
You need to log in before you can reply Login | Register now

Credit Rules