On the question of accuracy

[chenhaoGao-620]

Hello, I used WASS's software for wave reconstruction, but the results don't seem to be ideal, as shown in the attached image. Pls does wass pipeline have high requirements for image quality and how can I improve the quality of the results

[BlueGis]

Hello,have you solved the problem？

[fbergama]

Hello,
It is difficult to troubleshoot the problem just from the output. It looks like you are too close to the water surface, and therefore reflections make the stereo matching quite unreliable. Can you afford to put the cameras farther away?

Anyway, if you send me the full output of a stereo pair (for instance the 00000000_wd directory) I can help you sort out the problem.

[chenhaoGao-620]

Thanks for the reply.
The issue was indeed due to the camera being too close to the sea surface, resulting in a large matching error. When the camera was placed at a farther position, the effect improved significantly.

[chenhaoGao-620]

Thanks, I am interested in learning more about the measurement capabilities of the project. Could you please let me know the maximum range that your project can currently measure? Additionally, I would like to understand if there would be a significant measurement error if the camera is positioned too far from the sea surface, for example, at a distance greater than 300 meters.

Hello, It is difficult to troubleshoot the problem just from the output. It looks like you are too close to the water surface, and therefore reflections make the stereo matching quite unreliable. Can you afford to put the cameras farther away?

Anyway, if you send me the full output of a stereo pair (for instance the 00000000_wd directory) I can help you sort out the problem.

[fbergama]

In our typical installation, the cameras are placed roughly 5m apart to reconstruct an area of about 150 x 150 meters. You can scale things up or down depending on your specific needs. We had successful results when reconstructing areas of 700x700m with 20Mpix cameras, but that will be computationally expensive. Error is of course proportional to the distance.

[chenhaoGao-620]

In our typical installation, the cameras are placed roughly 5m apart to reconstruct an area of about 150 x 150 meters. You can scale things up or down depending on your specific needs. We had successful results when reconstructing areas of 700x700m with 20Mpix cameras, but that will be computationally expensive. Error is of course proportional to the distance.

Thank you very much for your response. Additionally, I would like to know the required baseline distance for measuring a 700m x 700m area. Based on the principles of stereo vision depth measurement, I have calculated that the depth measurement error caused by a 1-pixel disparity estimation error is proportional to (distance from the camera to the seasurface)^2/(baseline distance). So, if we want to ensure as accurate a measurement as possible for a 700m x 700m area, would a baseline distance close to 100 meters be necessary?

[fbergama]

700x700 is somehow extreme. Anyway, the stereo algorithm can match with subpixel accuracy, so approx 7-10 meters baseline would be ok

[Yuanzhang1245]

Thanks for the reply. The issue was indeed due to the camera being too close to the sea surface, resulting in a large matching error. When the camera was placed at a farther position, the effect improved significantly.

Hello, I also encountered this problem during my experiment.
Could you please tell me if the improvement was achieved by placing the camera further away or raising it higher?
Could you also provide me with the data on the shooting area and height for my reference? Thank you.