Has anyone else noticed the manual focus adjusting itself at random times? I still don't know what causes this and this video was just firing off consecutive photos and luckily caught it on video. Is this focus change expected or is this a problem that DJI needs to address. Since this is a photogrammetry focused camera a fixed focal length is critical. Trying to conduct a manual lens calibration I believe this issue is giving bad results.

There is the same problem with the M300 payloads. It's not really a Manual Focus, more like a semi random automatic focus. Another DJI fantasy.

LV_Forestry Posted at 2-24 23:39
There is the same problem with the M300 payloads. It's not really a Manual Focus, more like a semi random automatic focus. Another DJI fantasy.

Interesting. So how does DJI calculate the cameras distortion correction? Or is this value in the photos metadata the same for every M3E they manufacturer? They can't claim centimeter grade accuracy if the cameras focus is changing itself halfway through a mapping mission. For higher accuracy I'm using a pre calculated camera calibration in Metashape so when those same corrections get applied to the images of a different focal length that data becomes unusable. If what you're saying is true, this is unacceptable and DJI needs to address this immediately. It could be making everybody's data very inaccurate and they don't even know it.

Johnnokomis Posted at 2-25 09:57
Interesting. So how does DJI calculate the cameras distortion correction? Or is this value in the photos metadata the same for every M3E they manufacturer? They can't claim centimeter grade accuracy if the cameras focus is changing itself halfway through a mapping mission. For higher accuracy I'm using a pre calculated camera calibration in Metashape so when those same corrections get applied to the images of a different focal length that data becomes unusable. If what you're saying is true, this is unacceptable and DJI needs to address this immediately. It could be making everybody's data very inaccurate and they don't even know it.

Focus and focal length are two separate things.  
If you zoom, so if you change the focal length, in Metashape it will create different sensors classified by focal length.  No problem on that side.  

The focus is used to reposition the lens relative to the sensor to make the convergence point coincide.  This will only affect the sharpness of the image but not the dimensions.  

What bothers me with this manual focus, which isn't one, is that i set it to be as clear as possible on the treetops.  And this fool set it by himself on the ground.  I end up with sharp bits of map, then suddenly the treetops become blurry.

Metashape will discard all that is too blurry depending on your settings, there is no problem there.  The only solution I have found is to fly higher.  But suddenly if you need a lower GSD it's complicated.

For the accuracy of the map I'm not bothered because I import the point cloud from a LiDAR.  Then I let Metashape texture it according to the previously aligned images

Johnnokomis Posted at 2-25 09:57
Interesting. So how does DJI calculate the cameras distortion correction? Or is this value in the photos metadata the same for every M3E they manufacturer? They can't claim centimeter grade accuracy if the cameras focus is changing itself halfway through a mapping mission. For higher accuracy I'm using a pre calculated camera calibration in Metashape so when those same corrections get applied to the images of a different focal length that data becomes unusable. If what you're saying is true, this is unacceptable and DJI needs to address this immediately. It could be making everybody's data very inaccurate and they don't even know it.

For the distortion of the image, the parameters must be in the meta-data.  To find out if the M3E has this functionality, you add images in Metashape and without having performed the alignment you will see in camera calibration.  
If the parameters have a value it is because they are in the metadata, if they are zero...
If they are present, nothing prevents you from calibrating the camera with the Metashape checkerboard.  In order to have a personalized calibration.  I stopped doing it because I don't see a significant enough change.  I defer to the meta data provided by DJI.  
On the other hand for drones without this data, M2P, M2Z..., yes I do it systematically.

LV_Forestry Posted at 2-25 11:42
Focus and focal length are two separate things.  
If you zoom, so if you change the focal length, in Metashape it will create different sensors classified by focal length.  No problem on that side.  

"Does focusing affect the focal length?

Above we mention that focal length is related to focus distance.  Focal length is the principal distance of a camera when it is focused at infinity. In photogrammetry we are interested in the camera’s internal geometry at the time photos were taken – so it is the principal distance that we want to know precisely in photogrammetry.

The answer is ‘yes focusing a lens changes its principal distance’."

- Source: www.photomodeler.com/focal_length_4_dummies

I need focus to be on infinity and it never needs to leave that setting. Manual focus should not have any input other than a manual one.

Johnnokomis Posted at 2-25 12:02
"Does focusing affect the focal length?

Above we mention that focal length is related to focus distance.  Focal length is the principal distance of a camera when it is focused at infinity. In photogrammetry we are interested in the camera’s internal geometry at the time photos were taken – so it is the principal distance that we want to know precisely in photogrammetry.

They are funny at photomodeler.com, of course focus changes the main distance, that's how focus works.  

But the focusing lens remaining the same, the focal length does not change.  

Yes I agree that in manual mode it should not move.  But probably like for everything else we will never have explanations and we will have to live with it.

LV_Forestry Posted at 2-25 11:58
For the distortion of the image, the parameters must be in the meta-data.  To find out if the M3E has this functionality, you add images in Metashape and without having performed the alignment you will see in camera calibration.  
If the parameters have a value it is because they are in the metadata, if they are zero...
If they are present, nothing prevents you from calibrating the camera with the Metashape checkerboard.  In order to have a personalized calibration.  I stopped doing it because I don't see a significant enough change.  I defer to the meta data provided by DJI.  

I don't need to find out if it's there I already know it is and I can see the parameters that DJI is applying to the image. In a photo that had dewarping  turned off the values are in the metadata for programs to apply in post. My question was, are these values calculated for each M3E sensor individually or is this an averaged value that are used on every M3E they make?

I did similar to what Metashape's checkerboard calibration offers except on a way larger scale as shown here.. This screenshot is from a different camera's calibration but the process is the same.

If anybody else can chime in with their M3E's metadata that'd be great. It'll tell you the date it was calibrated and the corrections being applied. Here's mine - Dewarp Data : 2022-06-08 ;3713.290000000000,3713.290000000000,7.020000000000,-8.720000000000,-0.112575240000,0.014874430000,-0.000085720000,0.000000100000,-0.027064110000
I suspect others will have identical numbers.  

Johnnokomis Posted at 2-25 12:16
I don't need to find out if it's there I already know it is and I can see the parameters that DJI is applying to the image. In a photo that had dewarping  turned off the values are in the metadata for programs to apply in post. My question was, are these values calculated for each M3E sensor individually or is this an averaged value that are used on every M3E they make?

I did similar to what Metashape's checkerboard calibration offers except on a way larger scale as shown here.. This screenshot is from a different camera's calibration but the process is the same.[view_image]

I'm looking on the internet it's true that the focus definitely has an impact on the quality of pure photogrammetry.  
Using a LiDAR point cloud eliminates the problem.  

Theoretically the "f" setting should change.  So yes as you suppose it must be pre-recorded for infinity focus.  
But I find it hard to imagine how the distortion parameters can change because the lens always has the same shape.  Maybe I'm wrong.
No idea.  I will send an email to the DJI addresses I have.  We'll see.

Cool your checkerboard.  I always struggle with the screen. I never found the motivation to print one and keep it clean.

Johnnokomis Posted at 2-25 12:33
If anybody else can chime in with their M3E's metadata that'd be great. It'll tell you the date it was calibrated and the corrections being applied. Here's mine - Dewarp Data : 2022-06-08 ;3713.290000000000,3713.290000000000,7.020000000000,-8.720000000000,-0.112575240000,0.014874430000,-0.000085720000,0.000000100000,-0.027064110000
I suspect others will have identical numbers.

On P4M I am 100% sure that each drone has its own parameters. No idea for the others.

LV_Forestry Posted at 2-25 12:36
I'm looking on the internet it's true that the focus definitely has an impact on the quality of pure photogrammetry.  
Using a LiDAR point cloud eliminates the problem.  

The distortion should be consistent from Macro to Infinity, yes. No two camera lenses are the same however. My M3E is going to have a different correction value than the M3E beside it on the assembly line during manufacturing. This is expected and perfectly fine. What isn't acceptable is if DJI is baking in an averaged correction value across all M3E's while selling it as a survey grade measuring tool.

The (not so) simple workaround is to calibrate it yourself and use those corrections for every project. Problem is, the damn camera won't stay on infinity like it's being told to. I'm hoping this focus issue is specific to me because if this is widespread it should have been addressed long before now.

Johnnokomis Posted at 2-25 17:22
The distortion should be consistent from Macro to Infinity, yes. No two camera lenses are the same however. My M3E is going to have a different correction value than the M3E beside it on the assembly line during manufacturing. This is expected and perfectly fine. What isn't acceptable is if DJI is baking in an averaged correction value across all M3E's while selling it as a survey grade measuring tool.

The (not so) simple workaround is to calibrate it yourself and use those corrections for every project. Problem is, the damn camera won't stay on infinity like it's being told to. I'm hoping this focus issue is specific to me because if this is widespread it should have been addressed long before now.

Does the M3E have the option to "calibrate infinity focus" like the P1 does, and will it run infinity focus throughout the flight if you set that option in your mission settings as with the P1?

We are considering buying a few M3Es and this is very concerning. If it's consistent throughout the flight, Metashape can work with it. If it is changing throughout, toast.

Ryan H. Posted at 2-26 19:42
Does the M3E have the option to "calibrate infinity focus" like the P1 does, and will it run infinity focus throughout the flight if you set that option in your mission settings as with the P1?

We are considering buying a few M3Es and this is very concerning. If it's consistent throughout the flight, Metashape can work with it. If it is changing throughout, toast.

By default during a mission the camera is supposed to automatically go to infinity focus and stay there. Ever since I've started watching for it recently the focus has in fact stayed on infinity the whole time. It's during the manual calibrations taking pictures manually where the focus will change at random times, rendering the whole process useless. Metashape does a fairly decent job with their per-project based calibrations, but if you're trying to get every bit of accuracy possible a manual one is best.

No the M3E does not have a way to calibrate infinity focus like the P1.

Johnnokomis Posted at 2-25 12:16
I don't need to find out if it's there I already know it is and I can see the parameters that DJI is applying to the image. In a photo that had dewarping  turned off the values are in the metadata for programs to apply in post. My question was, are these values calculated for each M3E sensor individually or is this an averaged value that are used on every M3E they make?

I did similar to what Metashape's checkerboard calibration offers except on a way larger scale as shown here.. This screenshot is from a different camera's calibration but the process is the same.[view_image]

What are the physical dimensions of that board and what is it made of?

Ryan H. Posted at 2-27 11:30
What are the physical dimensions of that board and what is it made of?

It's 4' x 4' and made of thick glass with a vinyl graphic applied over it.  

Hey DJI, it has been a while since we've heard an update on this topic. Actually, we have never heard anything from you that this is even a known issue by your company. I'm here to let you know that it is still happening. I've seen it take place on 4 different M3E's now, so it's not an isolated bug. How about y'all actually address this..? This is a huge problem and nobody has even commented on it from DJI. Today I have video of it so there's no denying it.

The video is legendary!  
Unfortunately there will probably be no response.  Bots don't understand when there are too many words or when it's too technical.

LV_Forestry Posted at 5-10 20:14
The video is legendary!  
Unfortunately there will probably be no response.  Bots don't understand when there are too many words or when it's too technical.

Thanks! Maybe they know how to forward it to someone who can help them with the big words.

Submit this as a support ticket?  If we all do the same it must be noticed.

TonyPHX Posted at 5-11 05:06
Submit this as a support ticket?  If we all do the same it must be noticed.

They will answer you, apology please send it for waranty study case.  It will come back after 15 days with no defects found or they will replace it.  

Unless you have a backup, it is not conceivable to leave a drone for 15 days without much reason.  
A private user can do this, he'll just have a little pain for his hobby.  
But for a company it's technical unemployment and it hurts a lot.  

So as long as flight safety is not in question and we manage to find workarounds to overcome the problem, we content ourselves to go through routes that are supposed to be made for that.  But it is clear that we are not very well listened.

LV_Forestry Posted at 5-11 05:59
They will answer you, apology please send it for waranty study case.  It will come back after 15 days with no defects found or they will replace it.  

Unless you have a backup, it is not conceivable to leave a drone for 15 days without much reason.  

Sending it in will do no good. Like I said, this is the 4th M3E that I've witnessed this behavior on. In fact, the screen recording above was this M3E's very first flight, ever. I had updated it to the newest firmware less than an hour before this flight took place. The fact that DJI won't even acknowledge this major problem exist tells me they're aware of it but not enough to fix it.

LV_Forestry Posted at 5-11 05:59
They will answer you, apology please send it for waranty study case.  It will come back after 15 days with no defects found or they will replace it.  

Unless you have a backup, it is not conceivable to leave a drone for 15 days without much reason.  

Yes, I know you are right.  I'm just frustrated with this.  It has burned a few things for me.

I have emailed DJI Support Service EU regarding this issue and I received a reply that it has been forwarded and I should be contacted for a follow-up soon. We'll see.

Johnnokomis Posted at 5-12 07:09
I have emailed DJI Support Service EU regarding this issue and I received a reply that it has been forwarded and I should be contacted for a follow-up soon. We'll see.

Thank you for doing this, and I hope it leads to results.  Incidentally, it will do this odd focus behavior during video as well.  Really annoying to lose a nice orbit shot when the autofocus kicks in for a second.

TonyPHX Posted at 5-18 20:16
Thank you for doing this, and I hope it leads to results.  Incidentally, it will do this odd focus behavior during video as well.  Really annoying to lose a nice orbit shot when the autofocus kicks in for a second.

I wish I had better news to report from this but sadly I don't. I have a long email chain of me and 3 different DJI employees all discussing this focus issue. One person was puzzled by this behavior and requested the flight logs, which I provided. Another tried explaining how it has a fixed focal length and a shift in focus makes no difference. I finally got him to change his mind with this reply - " Yes, the lens has a nominal focal length, but the lens is not fixed focus. When you focus it, you invariably make small changes in the distance between the lens's center-of-projection and the sensor surface. While it's a small change to that nominal focal length, it has an effect beyond centimeter-level in Z when being used for photogrammetry. I suspect it also causes changes to the lens distortion parameters that are calculated as part of lens calibration."

After that the reason changed completely and the last email from them claims- "the position of camera infinity will change due to the temperature, so you can see the jump of focus on the APP, which is normal, but the physical position of focus has actually been set and has not changed."

This too makes no sense. Temperature does not make focus shift and nothing about this behavior is normal. I've proven my case to DJI but it doesn't seem to be going anywhere. I guess they've just accepted this flaw. Which is sad.

Johnnokomis Posted at 5-18 20:35
I wish I had better news to report from this but sadly I don't. I have a long email chain of me and 3 different DJI employees all discussing this focus issue. One person was puzzled by this behavior and requested the flight logs, which I provided. Another tried explaining how it has a fixed focal length and a shift in focus makes no difference. I finally got him to change his mind with this reply - " Yes, the lens has a nominal focal length, but the lens is not fixed focus. When you focus it, you invariably make small changes in the distance between the lens's center-of-projection and the sensor surface. While it's a small change to that nominal focal length, it has an effect beyond centimeter-level in Z when being used for photogrammetry. I suspect it also causes changes to the lens distortion parameters that are calculated as part of lens calibration."

After that the reason changed completely and the last email from them claims- "the position of camera infinity will change due to the temperature, so you can see the jump of focus on the APP, which is normal, but the physical position of focus has actually been set and has not changed."

Well, at least you are tried and I hope still trying.  I am going to pile on with them next with some examples of ruined orbit footage.

TonyPHX Posted at 5-19 13:01
Well, at least you are tried and I hope still trying.  I am going to pile on with them next with some examples of ruined orbit footage.

Here's DJI last response and I think my support case is finished. I can't prove my side of the argument because there's no metadata that holds the focus value, only one for focal length.

The range of the MF scroll bar of the camera is INF (the focus motor position corresponding to the hyperfocal distance) to MOD (the focus motor position corresponding to the closest focusable distance). Normally, the INF position changes with temperature. In MF mode, the focus distance (i.e. the focus motor position) set by the user on the APP is fixed. When the camera temperature continues to rise during the flight, the INF changes will cause the range of the MF scroll bar to change simultaneously, and for the user, it feels like the focus distance set on the APP has jumped. But in fact, the scroll bar can be set to change the range, and the focus distance set by the user is still valid.

Please rest assured that the focus does not affect the focal distance, the focus mode of the drone in the route is MF, which is the focus set to infinity.

Hello. We apologize for the inconvenience. The range of the scrollbar of the camera in MF mode is from INF (the focus motor’s position corresponding to the hyperfocal distance) to MOD (the focus motor’s position corresponding to the minimum available focus distance), and INF position changes with temperature.
  
In MF mode, the focus distance (i.e. focus motor’s position) set by users in the app is fixed. During flight, INF changes with the continuous rise in the camera’s temperature, and thus the range of the scrollbar of the camera in MF mode changes accordingly.

For pilots, it seems that the focus distance set in the app changes. Actually, the adjustable range of the scrollbar changes, while the focus distance set by pilots remains effective.

Johnnokomis Posted at 2-25 12:33
If anybody else can chime in with their M3E's metadata that'd be great. It'll tell you the date it was calibrated and the corrections being applied. Here's mine - Dewarp Data : 2022-06-08 ;3713.290000000000,3713.290000000000,7.020000000000,-8.720000000000,-0.112575240000,0.014874430000,-0.000085720000,0.000000100000,-0.027064110000
I suspect others will have identical numbers.

Confirmed.
2022-06-08;3713.290000000000,3713.290000000000,7.020000000000,-8.720000000000,-0.112575240000,0.014874430000,-0.000085720000,0.000000100000,-0.027064110000

Adam9997 Posted at 6-29 03:08
Confirmed.
2022-06-08;3713.290000000000,3713.290000000000,7.020000000000,-8.720000000000,-0.112575240000,0.014874430000,-0.000085720000,0.000000100000,-0.027064110000

Thank you sir for the confirmation. I'll give DJI a pass on this one since using these corrections are off by default. What's weird is that I remember with the Phantom 4 RTK, it was recommended to have this turned on during a mapping/modeling flight plan. Maybe these older drones received camera specific values..?