Please select Into the mobile phone version | Continue to access the computer ver.
OA3 not suitable for in-car-use? (horizon levelling)
1323 28 2023-5-15
Uploading and Loding Picture ...(0/1)
o(^-^)o
SeehawerB
lvl.4
Flight distance : 5414203 ft
  • >>>
Germany
Offline

I tried all stabilization modes (RS, RS+ and HS) and all share the same problem: if I use the OA3 to record a drivers view in my car the horizon tilts every time I drive around bends. Obviously the OA3 does not take the image content into account, relies only on its built in position sensor, which is wrong because any lateral forces lead to a tilted image.


How can I get stabilized footage that stays levelled?
My older Sony FDR-X3000R does this job perfectly with its (optical) stabilization. Never had this issue there.

I need stabilized but horizontal footage when recording inside moving cars. Is this possible?
2023-5-15
Use props
Iancraig10
lvl.4
  • >>>
United Kingdom
Offline

It does look quite weird because it doesn’t tilt ‘into’ the bend!

I’m guessing that this must be a problem with the way that levelling is done in the camera. Maybe try recalibrating the level, but I don’t think that’ll get it in truth. Levels just gets the horizon ….  That  is wandering and looks like an update is needed.
2023-5-15
Use props
johansenfoto
Second Officer
Flight distance : 1173379 ft
  • >>>
Norway
Offline

**Deleted**
2023-5-15
Use props
SeehawerB
lvl.4
Flight distance : 5414203 ft
  • >>>
Germany
Offline

Does not look weird to me, just wrong. The thing is, that obviously stabilisation in the OA3 only depends on where the sensor says is earths gravity pulling to. The moment you drive a bend the gravity and the centrifugal force superpose and the resulting force points away from the center of the curve you are driving.

With this behaviour it is not possible to get a stable horizon while driving in curves. The system is laggy so brief changes in direction do not show any reaction, but if you drive a curve for some seconds it happens in the way the video shows it. Try it out yourself.

@DJI What is the point of a Horizontal stabilisation that does not work if the camera encounters lateral forces? You cannot rely only on the acceleration sensor in the camera. This is a major flaw. Do I have the wrong settings or is the camera just not capable to keep the footage levelled?
2023-5-16
Use props
SeehawerB
lvl.4
Flight distance : 5414203 ft
  • >>>
Germany
Offline

Recalibration is no solution. The sensor is calibrated, you see it on the straights where the camera image is horizotal.
2023-5-16
Use props
SeehawerB
lvl.4
Flight distance : 5414203 ft
  • >>>
Germany
Offline

The OA3 DOES NOT do HorizonSteady/HorizonBalancing in a correct way. Although this is exactly the mode used in the video, you can see that it doesn't work
2023-5-16
Use props
Iancraig10
lvl.4
  • >>>
United Kingdom
Offline

SeehawerB Posted at 5-16 01:32
Does not look weird to me, just wrong.

It doesn’t even tilt 'into' the bends which is why it looks so peculiar.

I wonder whether dji could ‘reverse’ the stabilisation detection adjustment  to at least get the tilt going a more natural way if the design  means that it can’t actually be held still! At least it might look more natural then.
2023-5-16
Use props
SeehawerB
lvl.4
Flight distance : 5414203 ft
  • >>>
Germany
Offline

It just follows the vector of force, it levels to where it thinks is earth. But this is rubbish, you may not do this that way. Does not work.

If it says HorizonBalancing or HorizonSteady I expect the camera to do so. And though it calls itself "action camera" I expect it to do so moving.
2023-5-16
Use props
Iancraig10
lvl.4
  • >>>
United Kingdom
Offline

Yes, it looks like a design problem for sure.
2023-5-16
Use props
Fishycomics
Second Officer
United States
Offline

thank you. glad I am not the only one complaining on this but will follow
2023-5-16
Use props
johansenfoto
Second Officer
Flight distance : 1173379 ft
  • >>>
Norway
Offline

**Deleted**
2023-5-16
Use props
Fishycomics
Second Officer
United States
Offline

johansenfoto Posted at 5-16 04:56
Osmo Action 3 is perfect, stop complaining or go buy gopro

But serious, how else would horizon work? How does it work on other brands?

not me, I am following
2023-5-16
Use props
Robox
lvl.4
Flight distance : 343727 ft
Russia
Offline

At a speed higher than the response of the sensor, the opposite force acts on the sensor and it turns out that the sensor 'goes crazy'. The same is true if you shoot near the window of an aircraft during takeoff or landing acceleration.
2023-5-16
Use props
SeehawerB
lvl.4
Flight distance : 5414203 ft
  • >>>
Germany
Offline

@DJI
Do you have any solution or at least a comment on this weird behaiviour? Is it fixable?
2023-5-16
Use props
DJI Tony
Administrator

Offline

Hi, SeehawerB. Thanks for reaching out. We're sorry for the inconvenience this has caused. We have forwarded your concern to our engineers for feedback. We will keep you updated.
2023-5-17
Use props
SeehawerB
lvl.4
Flight distance : 5414203 ft
  • >>>
Germany
Offline

Thank you very much. I would highly appreciate if there was a fix for that. Right now the not working horizontal leveling is rendering about half the applications I use my 2 OA3s in useless.
2023-5-17
Use props
DJI Tony
Administrator

Offline

SeehawerB Posted at 5-17 00:05
Thank you very much. I would highly appreciate if there was a fix for that. Right now the not working horizontal leveling is rendering about half the applications I use my 2 OA3s in useless.

Hi there. Thanks for your patience. Here are the possible causes of the behavior:
1. Primary cause:
When the stabilization function is enabled in the Osmo Action series products, the devices will judge the direction of gravity based on the gyroscope in the IMU, and determine the horizontal angle of the picture afterwards.
However, the inertia caused by sudden acceleration/deceleration during driving or the centripetal force generated by turning will "offset" the gyroscope's judgment on the gravity direction, so under certain circumstances, the screen will be skewed in different directions.
2. Secondary cause:
During the driving process, the car will show a certain tilt in scenarios such as sudden acceleration, deceleration and turning, the Osmo Action Series product will also tilt along with the car because it is attached.
And as confirmed by our engineers, this is due to the limitation of the stabilization principle. If the movement of the vehicle you are in cannot be adjusted, it is recommended to turn off the stabilization function. We apologize for the inconvenience this has caused you.
2023-5-17
Use props
SeehawerB
lvl.4
Flight distance : 5414203 ft
  • >>>
Germany
Offline

Well, that's exactly what I was contemplating about. The OA3 relies only on the gravity sensor = IMU. And this is not good at all. Other OEMs do that better. And this should be, because mounting Action Cameras fixed on moving objects is a normal use case, it's the main use case. Take a car, or a drone, or a train, or ... think about it, in almost any case the camera does not follow the center of gravity within an external gimbal mount, in every case the camera is mounted fixed on an object, if the object changes the direction like on rails, the image will tilt LIKE A PENDULUM. This is not as this should work. Sorry, DJI. You should also take the image content into account. This would give you the immediate clue that the image must not be tilted.

What the OA3 needs is a stabilisation WITHOUT TILTING. Or even better, a horizontal stabilization THAT TAKES THE HORIZONT AS INPUT VALUE AND NOT THE IMU.

Big fail, DJI. Sorry.
2023-5-17
Use props
Iancraig10
lvl.4
  • >>>
United Kingdom
Offline

Does the GoPro use the same principal? I’ve never seen this before.
2023-5-17
Use props
SeehawerB
lvl.4
Flight distance : 5414203 ft
  • >>>
Germany
Offline

At least my two "old" Sony FDR-X3000R have a very good stabilizsation which works flawlessly, they do not tilt when the are driven around bends or corners. They use a partly opto-mechanical partly software mechanism. I had not suspected that today other brands are so much more behind the technical possilities. The Sony are more than 5 years old. The image quality is not as crisp as it is with DJI products but still good.
2023-5-17
Use props
SeehawerB
lvl.4
Flight distance : 5414203 ft
  • >>>
Germany
Offline

This is a video from inside my car with a Sony FDR-X3000R mounted in the same manner as the OA3 in the Video in the opening post:



As you can see, though the camera is mouted directly to the car, the stabilization really depends on the horizon / the image content, not the movement of the car. You see the car "work" with the road, it moves a bit, tilts left and right how the road surface goes, the Camera image stays focused on the landscape. THIS is what I did expect the DJI Osmo Action 3 to also do. Simple as that. And now, after investing in two new cameras I have to admit the stabilization is rubbish, cannot do the things my old cameras could do, simply film from inside a car. SHAME!
2023-5-17
Use props
Mgozer
lvl.4
Switzerland
Offline

I use the first Osmo Action an it works really well in my car, even with rock steady enabled.
I think the better solution could be to disable the rock steady and horizion levelling for the time being.
2023-5-18
Use props
SeehawerB
lvl.4
Flight distance : 5414203 ft
  • >>>
Germany
Offline

That's what I will try next. Disabling all stabilization modes. How much the image is shaken then if my car is driving on b-roads.
2023-5-18
Use props
Andreas Dück
lvl.4
Switzerland
Offline

SeehawerB Posted at 5-18 07:30
That's what I will try next. Disabling all stabilization modes. How much the image is shaken then if my car is driving on b-roads.

I've also noticed this problem. On the motorcycle I immediately gave up horizon balancing since the first test video I made. The result was better with normal image stabilization, but there was always a drift where the image was not horizontal, although the camera was aligned horizontally. I've looked at how the problem could be solved, but we quickly end up with the Kalman filter or even the EKF... Topics I'm not familiar with and can't influence. If the picture were always aligned with the horizon but not stabilized, then you could stabilize it afterwards, but the way it is now it's really difficult.
2023-5-18
Use props
osmonauta
lvl.4
Hungary
Offline

Starts at 1:45. Was posted 7 months ago.



2023-5-18
Use props
SeehawerB
lvl.4
Flight distance : 5414203 ft
  • >>>
Germany
Offline

Thanks for the video!

So, DJI made an Action Cam with a stabilization working correctly only if the action goes straight ahead. Avoid bends, do not turn, because, well, it's an Action Cam, but no Moving-all-around Cam...
2023-5-18
Use props
Homeyjay
lvl.3
Flight distance : 33606 ft
  • >>>
United Kingdom
Offline

DJI, maybe add a ‘Speed’ mode where tilt data is just reversed?
2023-5-22
Use props
SeehawerB
lvl.4
Flight distance : 5414203 ft
  • >>>
Germany
Offline

Osmo Action 4 has the same bad behaviour:


(jump to 9:55 to see the horizon issue)

I expected that because the OA4 seems so be a so minor upgrade that DJI will not have addressed that. And personally I think DJI doesn't care for this problem at all.
So I have to go back to Gopro for in-car-usage.
2023-8-3
Use props
Iancraig10
lvl.4
  • >>>
United Kingdom
Offline

Not great is it? I when you thing about, it’s really not a good idea to have horizon levelling based on gravitational forces! That means you can’t go around sharp turns without it happening and as that video shows, GoPro doesn’t do it.
2023-8-3
Use props
Advanced
You need to log in before you can reply Login | Register now

Credit Rules