Support Website Contact Support Blog

Ghosting in orthoplane

hi,

 

I’m running into problems with the orthoplane function for an orthofacade. 

There is a lot of ghosting and objects in the foreground visible, rendering the orthofacade unusable.

I have cleaned up the point cloud, regenerated the 3D mesh into a very nice model, but when I render the orthoplane it looks nothing like the clean 3D mesh. 

Is there anything else I can do? 

For confidentiality reasons I cannot post the big pictures here (willing to send them to support privately though), but find an sample below of part of the mesh and the orthofacade.

 

Can you help?

 

Thanks,

Pierre

 

Hi Pierre,

 

As discussed in the request, please make sure the orthoplane is as thin as possible and contains only the wall of the facade and not a big part of the building.

Also if your building has the facade on a different depth report an orthoplane for each wall instead of drawing a bigger one that includes all of it.

 

Cheers,

Ina

Hi,

I am running into similar issues with the ghosting and strange renderings of an orthoplane.

What I would like to be able to produce are side plans of the building to be compared to the original plans from the architect. This might be a tall order, but something I thought was possible through the Orthoplane function.

Below I have posted (1) the 3d mesh that shows a reasonably clean model, and the (2) the resulting orthoplane of this side of the building.

In my final product, I need to show the large facade with height up to the top beam of the roof and the facade of the extension perpendicular to the main house.

Any recommendations to obtain the required outputs are very welcome.

The project contains 270 images with NADIR, oblique 45deg., oblique 60deg. and terrestrial images. No GCPs but enough MTPs for all images to calibrate and the model to render correctly.

Thank you

Charles

 

I had similar issues. But when I make a flight with horizontal pictures of the facade, and use them to a new project, I have a 90% success rate of doing a good orto.

Here is one example

The area of interest its very good and precise in scale, as I control it with the point cloud

Thanks Jorge,

You are suggesting that the best way to get a good orthoplane is to take the imagery from the same desired angle.

I was afraid this would be the case. This article (https://community.pix4d.com/t/5934-Orthoplanes-using-a-combination-of-terrestrial-and-aerial-data) suggests that the algorithm that chooses the (best) image is heavily biased towards ray angle over proximity or GSD.

I probably won’t have time to re-run the project with vertical grid flights around the project site, but will try and do this next time although the number of pictures (and processing time) then increases significantly. Having said that, maybe having only one camera model in the final project then makes processing easier.

Thanks for your help.

Any other recommendations for getting the best without returning to the site?

Charles

The best recommendation I can give its that take loads of data on site, and in the office just choose the data that fit the better for your project. Take Nadir photos, round-a-point photos, the most possible angles that you can get safely, and then just choose the ones for you purpose.

If you got the time and the batteries of course.