Rolling shutter gain

What is the real gain using rolling shutter correction? In other words what outputs are affected by rolling shutter choice? I’m processing the same project with and without rolling shuette cirrection but i don’t notice any difference. I’m talking about a Phantom 4, not Pro one.

It is not necessary that every dataset needs to be processed with the rolling shutter algorithm option enabled. In general, default settings already give good results.

Some cases, high wind, high speed and/or low altitude of the drone can increase the linear shutter effect and enabling the option could produce better results. 

Linear rolling shutter correction can be directly reflected in the processing quality of Step 1 (better feature extraction and matching, lower GCPs errors in the quality report, … ) and in the results of all the next steps. 

More information: 

As a side note, to better assess the quality of the results we always recommend to introduce GCPs and checkpoints in the project.

Best,

Well, as you say “enabling the option could produce better results.” is an admission that it may be or not, isn’t it?

So my question is: how and where I can see the differences and the improvements? Are there numbers or report to check? And if yes, in which step or folder they are?

The Rolling Shutter Effect can only happen with a camera that has a Linear Rolling Shutter. However, this does not mean it will reflect in the results every time you are using a camera with a Linear Rolling Shutter because the effect generally only happens if the drone flies fast or at low altitude.

This is why the Linear Rolling Shutter is not selected by default in the camera model, because there would be no Rolling Shutter Effect to correct in most cases, which means it would not improve the outputs. This also explains why the default shutter type is named _Global Shutter or  Fast Readout _. There are a few tools you can use to know when the Rolling Shutter Effect happens. These will enable you to know when to select the Linear Rolling Shutter in the camera model options to compensate for it. The tools are: 

  1. The Vertical Pixel Displacement is higher than 2.
  2. Quality report
    There are sections in the quality report where you can find hints to apply the Linear Rolling Shutter to the camera model: 

a)  Geolocation details 

In some cases, if you have a camera with a Linear Rolling Shutter and you have processed the project with the Global Shutter or Fast Readout shutter type, there will be large differences in the camera optimization and the georeferencing. This could be explained by the Rolling Shutter Effect. Note: In some cases, it is due to a wrongly selected camera model. An example before/after below: 

Without the correction:

With correction:

b)  Bending of the scene
The Rolling Shutter Effect tends to bend the results. This effect can be noticed in the DSM preview in the quality report, it would look curved as in the screenshot:

3.) Tilted objects
In some rare cases, the effect of the rolling shutter can be reflected in tilted objects in the reconstructed scene, meaning that the objects are not vertical/ perpendicular to the ground. This can be seen while inspecting the results in rayCloud or third party software.

This article describes the Rolling Shutter Effect and when to apply the corrections in more detail: Rolling Shutter Correction article
At the end of the article, there is a link to download example quality reports, for commodity I have copied the link here: Quality Reports

Please let me know if you need additional information to understand when to apply the Linear Rolling Shutter type in the camera model options.  
Thank you.

1 Like

Thanks for your answer: this is what I call an exhaustive answer.

If I’ve well understood it depends: if Camera Optimization has bad results and Georeferencing too, it may be possible try using Linear Rolling Shutter correction in case of a drone equiped with a Global Shutter (when speed is high or/and altitude is low).
Of course if you are using a Phantom 4 Pro, whic is equiped with a mechanical shutter, no need occurs, isn’t it?

Exactly, if you see non-optimal results in the above-mentioned sections while using a drone equipped with a linear shutter camera (e.g. P4)  we would recommend considering Linear Rolling Shutter correction as a possible solution. 

Just a small note, Global shutter cameras acquire the image all at once and there is no need to use the Linear Rolling Shutter correction.

More information about the Phantom 4 Pro sensor and shutter can be found in this community post: Phantom 4 Pro Rolling shutter effect.

Best,

“Just a small note, Global shutter cameras acquire the image all at once and there is no need to use the Linear Rolling Shutter correction.”: it means that for a Phantom 4 Pro, equipped with a mechanical shutter, no need occurs, isn’t it?

For more information about Phantom 4 Pro sensor and shutter and in order to avoid duplicating content in the community, we would like to invite you to read the Phantom 4 Pro Rolling shutter effect post for more information. 

Done, thanks.

 

Hi. I’m experimenting Pix4D demo for mapping and pile measurement purposes. I use a DJI M2P with a rolling shutter and I have questions about the way to manage rolling shutter compensation.

Q1 : I did a test flight in Pix4D Capture safe mode. I normaly don’t have any rolling shutter effect in this mode as the drone stops for each photo, but when trying to activate compensation (for test purpose), I did see correction applied, camera speed detection, and I don’t understand why (see image below). Did Pix4D estimated the drone’s speed with time and position and did a big mistake by ignoring that the drone stoped at each photo?

Q2 : Will I get same quality in safe mode / no RS compensation and high speed with RS compensation? I wonder if I’d rather fly safe with lower overlaps or fast with RS compensation and higher overlaps. In other words, is RS compensation almost as accurate as using a global shutter camera?

Thanks for your advises.