Support Website Contact Support Blog

Problems in results of Data Processing.

Dear,

We have collected data of a particular place that have various topography and Buildings at altitude 100mtrs. Sidelap 65% frontlap 75%. using Phantom 4 Pro

After processing the data we got an excellent quality report.(attached within post).

But the output ortho map and DSM shows distorted oblique and Bend buildings.

 

 

Please help us with the same and let us know the exact problem.

Hi, 

Thanks for sending the entire quality report and log file through a support request. Here are a few points that affected the outputs and that I would consider to improve the project: 

  1. Overlap
    In the quality report, it seems as if there was wind while you acquired the images, the overlap was not constant higher in one direction than the other: 
    This on the other hand had an impact on the overlap in the project, which can influence the reconstruction (yellow and red is insufficient overlap):

    For your next missions, if you notice a strong wind, I would increase the overlap, so that even when the drone flies in the direction of the wind the overlap is sufficient. 

  2. Driver error

The log file shows this error: 

“[Error]: Error e0094: It was not possible to initialize OpenGL. Verify that you have the latest drivers for your graphics card and that you are not using the software from a remote desktop. Detected QGLFormat::OpenGLVersionFlags: 2048”

I would try to update my GPU drivers and read this article: https://support.pix4d.com/hc/en-us/articles/202561189 

  1. Step 2 not processed

In the log file, it seems like step 2 was not processed. When that is the case, the DSM and Orthomosaic are based on the outputs from step 1, which is a sparse point cloud. This could explain the buildings that have “blobs” on them. The results should be more accurate if you process step 2. Point Cloud and Mesh. 

If you process step 2, you will have a densified point cloud, which will be more accurate and that can also be edited to remove noise, see: https://support.pix4d.com/hc/en-us/articles/202560499

  1. Add Manual Tie Points (MTPs)

There seems to be noise around buildings, e.g. the white roof in one of your screenshots. These can be tricky to reconstruct as the entire roof looks the same (all white), so it is difficult to find similar keypoints between images and thus correctly reconstruct the roof. 

For such areas, I would suggest to add MTPs in the problematic areas (and reoptimize): https://support.pix4d.com/hc/en-us/articles/202560349

  1. Hardware

I noticed that you process a rather large project on a small machine. If that would be an option, I would suggest to upgrade the hardware to improve the processing speed. More in these articles: 

Hope this helps! Let us know if it improved the project. :slight_smile:

I have a question about one of your comments.

"In the log file, it seems like step 2 was not processed. When that is the case, the DSM and Orthomosaic are based on the outputs from step 1, which is a sparse point cloud. This could explain the buildings that have “blobs” on them. The results should be more accurate if you process step 2. Point Cloud and Mesh. "

So my question is about weather I need to process the mesh?  If I do not need the mesh as an output, do i still need to create it to help create a better orthomosaic?  I do create a point cloud every time and then the Orthomosaic, but should I be creating a mesh as well to make a better results for the Orthomosaic.  Also does that go for the DSM as well and should I be creating a DSM to help with Orthomosaic?
I did not think I needed to do any of that, just the Point Cloud, so I would like some clarification on this process to ensure I am getting the best results that I can.  

Thanks,

Anthony

@Anthony: Good question! No need to create the Mesh, the creation of the Mesh can be deactivated in the processing options by unchecking the box “Generate 3D Textured Mesh” under Process > Processing Options > 2. Point Cloud and Mesh > 3D Textured Mesh > Generation as seen below: 


This will speed up the processing as well, as less outputs are created. 

The reason why the Mesh is not necessary in this case is that the Orthomosaic is based on the Digital Surface Model (DSM), which itself is based on the point cloud. The software will use the densified point cloud generated in step 2 if it was created, if not, it will use the sparse point cloud from step 1. Hence, the Mesh is not necessary. 

As mentioned in the explanation above, the Orthomosaic is based on the DSM, so yes, you have to create the DSM. 

Let me know if there are questions remaining! :slight_smile: