Hello,
I recently used a drone to capture an extensive industrial area ex) over 5KM x 5KM. The captured data was exported as a .json file, including GPS values, and uploaded to Cesium Ion. However, some buildings appeared blurry. When capturing additional images, I used a mix of DSLR and drone photography.
When reconstructing only the newly captured area as a new project, the existing optimal texel size was 0.02, but the texel size of the reconstructed model was 0.006. The reason for this choice was that the new area had about 1,700 photos, while the previous area had about 13,000 photos, and due to heavy vehicle movement, I was concerned that the alignment might not work properly.
Reality Capture had issues where the position changed even when aligning the same area multiple times using GPS. Additionally, when trying to merge meshes created from separate new projects, I encountered difficulties due to misalignment.
I have a few questions related to this:
-
When aligning an extensive area, there are cases where the boundaries do not match even when using GPS across different projects. Is there a way to resolve this?
-
If part of the extensive area has low quality and I want to work with additional photos, what methods can I use?
-
Is there a way to change the optimal texel size?
I would appreciate your advice. Thank you!
Hello @byeoung_jun
The optimal texel size is not related to the number of captured images. It is basically a GSD and it is related to the capturing distance from the object and used camera.
What kind of drone are you using? Do you have a RTK drone?
1. When aligning an extensive area, there are cases where the boundaries do not match even when using GPS across different projects. Is there a way to resolve this?
The problem could be, that the drone’s GPS is not so precise and sometimes it could be quite different for various flights. The best option there is to use RTK drone or measured GCPs placed on the ground.
2. If part of the extensive area has low quality and I want to work with additional photos, what methods can I use?
You can use the same method as before also with adding more detailed images.
3. Is there a way to change the optimal texel size?
The optimal texel size can be change with the flight height/distance to the captured object. There is also the option to set the texel size as Custom.
Thank you for your response! @OndrejTrhan
We are using the Mavic Enterprise 3 RTK Drone.
Regarding point number 2, is it correct to proceed with the alignment by adding the additional photos taken?
Additionally, we need to carry out a project over a very large area of more than 10 km x 10 km. Previously, when we attempted to align data with over 60,000 photos, we encountered issues because it exceeded the limit of 60,000 photos. Do you have any good solutions for this?
Are these RTK coordinates implemented to the images? What are your alignment settings? Do you consider the achieved precisions of the images in your process?
Yes, it is basically a normal process for such situation. Just try to capture the images in similar conditions (like cloudy one).
There is no image limitation. The only limitation is your HW. If the number of the images is over the RAM limitation, you can set the lower number for detected features in the alignment settings.
Thank you very much. @OndrejTrhan
Thank you very much. We are attaching our files that we replaced for the alignment settings and image accuracy. Additionally, we are using the EPSG:4326 - GPS (WGS 84) coordinate system. Furthermore, it’s challenging to capture a wide area in just one day, so we tried to shoot under the same conditions over several days. However, during the alignment process, the vehicle’s position and count can vary each day, which makes it difficult to adjust over a large area. Do you have any good suggestions
As you camera has 20 Mpix, it is enough to use Max features per images set to 40 000 (https://www.youtube.com/watch?v=xMeW_ZuCa_I), also, you can set it accordingly to your image number and RAM or you can process the data in the increments and then align them all together.
As you are using RTK, you can also set the higher XYZ accuracy in the Camera prior settings (set it accordingly to your data). I suppose it should be around 3 cm, as you used RTK.
Vehicles are not so important in this case, the building have a bigger weight there. But yes, there could be some influence. The best way will be mask such areas out. You can also use measured GCPs around the area or you can add control points to align the data properly.
Thank you so much for helping me out!
Everything is organized now. You’re truly the best! If I have any more questions in the future, I’ll definitely reach out.
