How to solve strange images aligment

[Daniele]

Dear all,
I would like to hear if someone can help in understanding this strange photo alignment and maybe find a workaround to solve the problem.
During ROV monitoring, I am acquiring many underwater images along a pipeline with a GoPro action camera (Hero 7 in wide FOV). The images have a good overlap because they are acquired every second.
However, after the alignment process, the images are correctly aligned, but the shape of the sparse point cloud seems a spiral (screen in the attachment). Of course the pipeline is a straight line.
Do you have any idea about this strange behaviour?

Many thanks
Daniele

[JMR]

Precalibrate your camera could be the best solution.

[Daniele]

Precalibrate your camera could be the best solution.

Thanks, JMR,
for calibration, do you mean acquiring photos of the using checkboard and then estimating camera parameters in Agisoft lens?

[Alexey Pasumansky]

Hello Daniele,

Can you please also share the screenshots of Initial and Adjusted tabs from the Camera Calibration dialog and also specify the parameters used for Align Photos operation.

[Daniele]

Hello Daniele,

Can you please also share the screenshots of the Initial and Adjusted tabs from the Camera Calibration dialogue and also specify the parameters used for Align Photos operation.

Hi Alexey,
thank you for taking the time to read this post and helping me again. Your request has unlocked my mind to look probably towards the right direction.

 In fact, on the camera calibration tab, if I set  "auto" in the initial parameters tab, the focus length seems a very odd number (1733.43811), and after image alignment, the adjusted tab remains greyed out. Of course, I get my spirally sparse point cloud (figure 1)

However, if I chose the precalibrated option and manually set the focal length to 3 mm, the process led to better results because, after image alignment, I can see my images correctly aligned in a straight point cloud representing my pipeline. However, the adjusted tab still remains greyed out and also the elevation seems disappeared because the sparse point cloud is without the z component (Figure 2).

Do you have some suggestions to solve this situation?

[Alexey Pasumansky]

Hello Daniele,

Setting "3" as focal length value in Initial tab is not correct, as the value should be in pixels, not in mm, so I suggest to switch it back to Auto option.

However, which value do you get in the Adjusted tab after alignment? Also you may try switching the camera type from Frame to Fisheye, if the images have obvious fisheye distortions.

[Daniele]

Hi Alexey,
I switched to auto and set camera type to fisheye, but the generated point cloud with this setting is more strange than before, without any definable shape.
I am attaching the images with the adjusted parameters after both the alignment (fisheye and frame).

[Paulo]

hello Daniele,

i am wondering at what speed is your ROV moving? and what is the exposure time on your pictures? It is possible that the GoPro has a rolling shutter and maybe such compensation could help?

[Daniele]

hello Daniele,

i am wondering at what speed is your ROV moving? and what is the exposure time on your pictures? It is possible that the GoPro has a rolling shutter and maybe such compensation could help?

Hello Paulo,
thanks for your suggestion and for sure the GoPro as a rolling shutter but the speed of the rov was very slow less than 1 m/sec. I tried with that option but that compensation the results remain the same.