Rounded edges near roof, even when I use external perfect point cloud

Hi there,

I am trying to produce some orthomosaic, but I have some problemes near roof edges. For example, I have produced a perfect DSM using 3rd party software that look like this:

Because it is not possible to import DSM directly to Pix4D and produce ortho using my own DSM, I need to convert above DSM to point cloud and import this point cloud to Pix4D. Converting a raster DSM to point cloud is not really difficult using for example GlobalMapper, The problem is when Pix4D tries to reconvert point cloud to its usable DSM, it rounds edges, something like this:

And this DSM causes to produce a bad ortho. like this:


As you can see there are some movements near bottom edges, and some rounded edges around it.

Anybody any solution?

And I forget to mention, I have tried both Use Surface Smoothing and not using it, the result is the same:

Could you give more details on the source of which is the external DSM?

Because of such quality, I would expect to be from high-quality LIDAR data; otherwise, I would check if it was not cleaned or modified.

Also when extracting the points from it are the perfect edges preserved? Is the external point cloud aligned with the one created by Pix4D desktop? Could you post a screenshot? 

Thanks,

Ina

 

 

I am creating my first house project and I have problems with the reflection of the roof. I need to paint a self-adhered modified bitumen roof system. I can’t understand how to convey the inequalities of such a roof.

Does anyone have experience of such work?

Hi,

I am not sure I understood well your question: is the goal of your project to calculate the surface of the roof to know how much paint you will need?
Is Pix4Dmapper is failing to generate an accurate representation of your roof? This would probably be caused by a combination of challenging image content, because the roof reflects the sunlight, and a lack of overlap. If you can go capture images again, I recommend you go when the sky is completely clouded to avoid reflections and ensure you capture high image overlap, at least 85% frontal and 70% side overlap.

If I misunderstood your question or issue, please do not hesitate to provide more details.

Rhéa