DJI Mavic, Air and Mini Drones
Friendly, Helpful & Knowledgeable Community
Join Us Now

Mini 2 Effective field of view in different modes?

PaulWilliams

Member
Joined
Feb 6, 2021
Messages
10
Reactions
2
Age
56
Location
New Zealand
Hello group, first time poster.

Does anyone happen to have information on (actually in-image) field of view (FOV) for the video and photo modes of the Mini 2?

I know this has been partially answered elsewhere, at least for the Mavic Pro and P2P.

What I am really looking for is FOV in the horizontal and vertical for each mode. I have started to measure these myself, but it would be good to have confirmation from an independent test, or better still from first principals given spec's of the components and image parameters.

happy Droning...
 
Partial answer to my own question. I have been digging through the EXIF data of a few image files (with the use of "exiftool") and have discovered the Field of View is recorded. For the 4000x2250 pictures I am looking at exiftool reports " Field Of View : 73.7 deg" also of use "Focal Length : 4.5 mm (35 mm equivalent: 24.0 mm)"

So I thought I would dig further. The author of exiftool also wrote some C++ code that provide a little more detail at the EXIF data. It seems the 73.7 value of the Field of View is rounded from a value of 73.7398575770811 not immediately a very striking number, but read on...

Given other entries in this forum say the way to calculate the FOV is: FOV = 2 arctan(SensorSize/2f) and given I have a value for FOV and f (focal length), I thought I would see what I can find. Well :
tan( 73.7398575770811 / 2) = 0.750000851 , i.e. 0.75 to 6 decimal places.
Other forum entries suggest photographs are cropped from the photo-sensor so multiplying the above by 2 x f and seeing the suggested SensorSize of 6.75 is probably bogus.

Does anyone know why the very precise value of 0.75 falls out?
 
It looks like your calculations are correct. 6.75 is for a horizontal dimension of a drone sensor and it's in mm. First of all, I wolud not believe all data stored in the EXIF, especially in the DJI's EXIF. It's full of ****. I do a lot of exif data checks/manipulations and DJI is at the bottom of "trustworthy" companies in this regard, eg. every model I've already checked saves data with mistakes (shifts/wrong zone/mixed time zones/so on). But if we take given FOV for granted, your 6.75mm is close (but only close) to a theoretical size of Mini2's sensor which they say is 1/2.3" (ie. 6.30mm x 4.7mm OR 6.17mm x 4.55mm depending on a manufacturer). On the other hand if we take 6.75mm literally it would give us size of a (rather unseen) 1/1.9" sensor.
I'd say they simply put wrong data in the EXIF (why not, it's DJI) - there are too many variables and not too many certain values.

For (I guess) 6.3mm x 4.7mm sensor and 4.5mm of focal length it gives us FOV 70x55 deg (82 deg diagonally)
-> Camera Field of View Calculator (FoV)
 
Last edited:
  • Like
Reactions: PaulWilliams
Welcome to the forum and starting out with some heavy Questions.
 
It looks like your calculations are correct. 6.75 is for a horizontal dimension of a drone sensor and it's in mm. First of all, I wolud not believe all data stored in the EXIF, especially in the DJI's EXIF. It's full of ****. I do a lot of exif data checks/manipulations and DJI is at the bottom of "trustworthy" companies in this regard, eg. every model I've already checked saves data with mistakes (shifts/wrong zone/mixed time zones/so on). But if we take given FOV for granted, your 6.75mm is close (but only close) to a theoretical size of Mini2's sensor which they say is 1/2.3" (ie. 6.30mm x 4.7mm OR 6.17mm x 4.55mm depending on a manufacturer). On the other hand if we take 6.75mm literally it would give us size of a (rather unseen) 1/1.9" sensor.
I'd say they simply put wrong data in the EXIF (why not, it's DJI) - there are too many variables and not too many certain values.

For (I guess) 6.3mm x 4.7mm sensor and 4.5mm of focal length it gives us FOV 70x55 deg (82 deg diagonally)
-> Camera Field of View Calculator (FoV)
Tibzee,
Thanks for letting me know about the unreliable information in the EXIF. Do you happen to know if the GPS and yaw values have been accurate in DJI EXIF data you have looked at previously?

I have measured the field of view within an image (all be it not very accurately yet) to 65x40 (72 diagonally) for the 4:3 image mode. As mentioned elsewhere, they are probably cropping the sensor image down to this. I guess the cropping allows for some digital image stability to augment drone and gimbal stability.

I shall stay calm and carry on testing. And report back on any findings.

Regards
 
I have measured the field of view within an image (all be it not very accurately yet) to 65x40 (72 diagonally) for the 4:3 image mode. As mentioned elsewhere, they are probably cropping the sensor image down to this.
Usually it's the opposite, the sensor being 4/3 and being cropped for 16/9. Easy to find out by taking a photo in 4/3, one in 16/9 and a bit of video in each mode with the drone at the same location.
 
Tibzee,
Thanks for letting me know about the unreliable information in the EXIF. Do you happen to know if the GPS and yaw values have been accurate in DJI EXIF data you have looked at previously?

I have measured the field of view within an image (all be it not very accurately yet) to 65x40 (72 diagonally) for the 4:3 image mode. As mentioned elsewhere, they are probably cropping the sensor image down to this. I guess the cropping allows for some digital image stability to augment drone and gimbal stability.

I shall stay calm and carry on testing. And report back on any findings.

Regards
Hi,
so far GPS values always looked ok, ie position on the map was correct, with some marginal errors of course. I've never checked accuracy of yaw values, but it might be interesting to check it. About cropping: I haven't heard anything about it in DJI's drones, moreover we get 4000x3000px photos so it looks like a full sensor. What you see is what you get ;) I guess all stabilization is done via gimbal, no software stabilization via cropping.
BTW, panorama is made of 6x4 photos + 1 down + 1 starting, reference photo. So 360deg / 6 = 60 deg. And we need some overlap to stitch it well. Your calculations that FOV is 65x40 deg seems a little bit conservative, because to my eyes we have more overlapping. But it needs checking, my assumptions might be incorrect.

Cheers.
 
Last edited:
Hi,
so far GPS values always looked ok, ie position on the map was correct, with some marginal errors of course. I've never checked accuracy of yaw values, but it might be interesting to check it. About cropping: I haven't heard anything about it in DJI's drones, moreover we get 4000x3000px photos so it looks like a full sensor. What you see is what you get ;) I guess all stabilization is done via gimbal, no software stabilization via cropping.
BTW, panorama is made of 6x4 photos + 1 down + 1 starting, reference photo. So 360deg / 6 = 60 deg. And we need some overlap to stitch it well. Your calculations that FOV is 65x40 deg seems a little bit conservative, because to my eyes we have more overlapping. But it needs checking, my assumptions might be incorrect.

Cheers.
I have only just realised that the GPS values may be correct, but they are not very precise. I was just doing some tests to check the Yaw values by rotating the Mini 2 around me. But I have found I was not moving the drone far enough to change the GPS location values. I have just looking into the DJIFlightRecord_date_[time].txt files to find a more accurate GPS location. Of now I have a new issue that the timestamp in the JPG files does not seem to be quite synchronous to the DJIFlightRecord time (disregarding a local time vs UTC time offset).

On the up-side, I am impressed how stable this little drone is with just mechanical stabilisation.
 
Just being curious, is there any practical use of the FOV numbers in high precision ?
Yes.

It allows one (to some degree) to map an image (live, photo or video) to the real world, be that for image recognition, tracking, measuring, simulating, etc.
-There are a number of discussions in these groups about knowing FOV so simulated flights with google earth/streetview can match the images of a live flight.
- Given the lack of collision avoidance, one might write software to measure the growing scale of an image to detect when an object is getting close (the FOV is needed to translate the change in image scale to the change in range to an item).
So knowing the FOV is practical for several uses.
 
  • Like
Reactions: boblui
I have only just realised that the GPS values may be correct, but they are not very precise. I was just doing some tests to check the Yaw values by rotating the Mini 2 around me. But I have found I was not moving the drone far enough to change the GPS location values. I have just looking into the DJIFlightRecord_date_[time].txt files to find a more accurate GPS location. Of now I have a new issue that the timestamp in the JPG files does not seem to be quite synchronous to the DJIFlightRecord time (disregarding a local time vs UTC time offset).

On the up-side, I am impressed how stable this little drone is with just mechanical stabilisation.
Hi,
I don't have my notes with me right now but AFAIR all DJI's devices a tested (ie. Mavics and Osmos) have glitches with precise time. I took photos/videos of a atomic time clock and videos were/are saved using ZULU/GMT time zone, JPGs/DNGs were/are in a local time but with time shifts (constant time difference from the correct time) and every model has it's own agenda and might be even firmware version related. I have to double check it. Since you cannot anywhere set the time manually, unlike other devices from other well-known brands, we depend on a DJI's internal "auto" logic which is far from perfect. Moreover, some models save in the EXIF the start time of the video, some the end time. And it applies both to the EXIF and file creation/modification time. It's just a big mess.

I talked to a DJI's support, "we'll send your valuable ideas to our technical team" and it was ~ 2 years ago (for the first time). But I was just spitting in the wind.
Eventually, I just added a couple of lines of a code to my app, now it corrects the time of the files.
I wonder if they care about everything else like they care about the time and data, how it all even works.

Cheers.
 
I have only just realised that the GPS values may be correct, but they are not very precise. I was just doing some tests to check the Yaw values by rotating the Mini 2 around me. But I have found I was not moving the drone far enough to change the GPS location values. I have just looking into the DJIFlightRecord_date_[time].txt files to find a more accurate GPS location. Of now I have a new issue that the timestamp in the JPG files does not seem to be quite synchronous to the DJIFlightRecord time (disregarding a local time vs UTC time offset).

On the up-side, I am impressed how stable this little drone is with just mechanical stabilisation.
CORRECTION.. I was wrong! The GPS values stored in the EXIF data in the images is precise. I was incorrectly using a tool to read them.
I was using the (very capable) 'ExifTool' from Phil Harvey which by default reports GPS locations to 2 decimals of seconds. If I had only read the documentation I would have known there is a parameter to request higher precision (if it exists in the data). The Exif data in the image is stored to 4 decimal places of seconds.

Well, I guess I learned a few things about the DJI log files when I was trying find a better source of location information.

I would now like to do the same for video files.
 
I have only just realised that the GPS values may be correct, but they are not very precise. I was just doing some tests to check the Yaw values by rotating the Mini 2 around me. But I have found I was not moving the drone far enough to change the GPS location values. I have just looking into the DJIFlightRecord_date_[time].txt files to find a more accurate GPS location. Of now I have a new issue that the timestamp in the JPG files does not seem to be quite synchronous to the DJIFlightRecord time (disregarding a local time vs UTC time offset).

On the up-side, I am impressed how stable this little drone is with just mechanical stabilisation.
CORRECTION.. I was wrong! The GPS values stored in the EXIF data in the images is precise. I was incorrectly using a tool to read them.
I was using the (very capable) 'ExifTool' from Phil Harvey which by default reports GPS locations to 2 decimals of seconds. If I had only read the documentation I would have known there is a parameter to request higher precision (if it exists in the data). The Exif data in the image is stored to 4 decimal places of seconds.

Well, I guess I learned a few things about the DJI log files when I was trying find a better source of location information.

I would now like to do the same for video files.
Hi,
I don't have my notes with me right now but AFAIR all DJI's devices a tested (ie. Mavics and Osmos) have glitches with precise time. I took photos/videos of a atomic time clock and videos were/are saved using ZULU/GMT time zone, JPGs/DNGs were/are in a local time but with time shifts (constant time difference from the correct time) and every model has it's own agenda and might be even firmware version related. I have to double check it. Since you cannot anywhere set the time manually, unlike other devices from other well-known brands, we depend on a DJI's internal "auto" logic which is far from perfect. Moreover, some models save in the EXIF the start time of the video, some the end time. And it applies both to the EXIF and file creation/modification time. It's just a big mess.

I talked to a DJI's support, "we'll send your valuable ideas to our technical team" and it was ~ 2 years ago (for the first time). But I was just spitting in the wind.
Eventually, I just added a couple of lines of a code to my app, now it corrects the time of the files.
I wonder if they care about everything else like they care about the time and data, how it all even works.

Cheers.
I guess sorting this out for hobby drones is not going to make them much money. But is must be better in the professional high end as they are being used for some serious survey work. I wonder if that is the issue... if you make the low end too good, you lose the high end market. That said, I am finding the Mini 2 very capable and now I am working around some of these issue with meta-data, it is look like quite powerful tool .
 
I guess sorting this out for hobby drones is not going to make them much money. But is must be better in the professional high end as they are being used for some serious survey work. I wonder if that is the issue... if you make the low end too good, you lose the high end market. That said, I am finding the Mini 2 very capable and now I am working around some of these issue with meta-data, it is look like quite powerful tool .
Well, I have an access, privately and professionally, to a few dozens of consumer and pro market devices, and while glitches sometimes happen, I can assure you - that kind of a mess is a DJI's specialty. Even, if you buy low-end camera from any big photo company (just first letters from their names: C, F, N, O, P, S) or most famous action camera brand (G) you'll get proper metadata and time. So, if I can get proper metadata in $200 P&S camera or $300 action cam, why not from $600 drone?
 
Uum I am possibly complete up a Gum Tree with this but what it you video or photograph two mutually perpendicular tape measures from two accurately set distances? Two distances to perhaps allow for inaccuracies in guessing where the plain of the sensor is
 
Uum I am possibly complete up a Gum Tree with this but what it you video or photograph two mutually perpendicular tape measures from two accurately set distances? Two distances to perhaps allow for inaccuracies in guessing where the plain of the sensor is
Then we are up the same Gum Tree. That is exactly the method I used for my numbers above.
 
Lycus Tech Mavic Air 3 Case

DJI Drone Deals

New Threads

Forum statistics

Threads
130,600
Messages
1,554,275
Members
159,607
Latest member
Schmidteh121