Hello I am using Capturing Reality to work on Surveying projects and I am becoming unstuck.The people I am working for usually use GPS pole surveys and they give the data to the client the next day.
The problem I am having is the flights are quicker but the processing time is taking around 4 Hours but then if anything goes wrong and I have to run it then my next day has gone.
I need to make the Point cloud really for exporting it to another software and I do use the 3d Render to create a backcloth for when I am working on my Volume data.
I have 32gb Ram
Intel I7 8700 3.2 Ghz
Nvidia Geforce 1070 TI 8 Gb GPU Ram
Samsung Evo Pro 1tb M.2 Disk with capturing reality installed on this.
I have changed my settings to preview mode and although it was much quicker it then had so many holes in my data I had to spend twice as long in my other software filling them in with new triangles to give me the correct Volumes from the Stock piles that I am measuring.
I’m not really sure what settings I can change to take the Preview quality better (an in-between preview and Normal) I normally use Normal quality and it is certainly good enough for me. but still a decent quality to not have holes in the created image.
I have tried to export the Point cloud and in Capturing reality it shows colours but then I export it to a .xyz file to the Apps In Cadd N4ce software and then it doesn’t take the colours in.
The Images I am using are from a DJI Phantom 4 Pro using the Pix4d capture app to take the images.
I then bring the images in and then use my Ground Control Points for the actual GPS data and ignore in Capturing Reality.
the Image alignment is really quick so if I could get the point cloud out with colours that would be brilliant. But if I then have to go through the full processing that’s great but If I could get better than preview but less then Normal that I guess would save on time there for the processing.
It then takes me another 3-4 hours work in the next software so late nights after a flight and processing leaves me up all night to get the data to the client the next day before 7am.
Use normal detail reconstruction and adjust the minimal distance between two vertices parameter (the higher the number the less polygons will be created and it should be quicker as well) Try 0.01 for example and move patiently above and above until you are happy with the fast pace and quality. Also, use image downscale parameter to define how many times will be images be downscaled in resolution for reconstruction process to make it even faster. Use reconstruction region wisely to “cut off” the parts you don’t need before reconstructing. To get the color information of your point cloud you need to use Coloring process on the final model and then after that you export your dense cloud.
Thank you very much for the response I am just going to give that a Try. I also have seen the other settings re. Detail Decimation Factor and "Maximum Vertex Count per Part should I change anything in there at all to make the processing quicker ?
I have just been looking at the "Image DownScale (Currently Set at 4) should I lower this value or make it higher to scale less times or more times to make it work on a more scaled down image. I am sorry if it is a silly question.
I see so many different settings but I havn’t found anywhere that I can really find what most of them do.
Detail decimation factor will lower the quality of the resulting model and speed up the process
I would not touch the maximum vertex count per part to the positive side in this particular case only lower if so, but this can be dangerous.
If your downscale is 4 going above will decrease the quality and going lower will increase of course. I would surely not go lower than 4, but your megapixel count matters. If you have 40mpx camera and you decide to go for 4, you should still have quite good detail. Doing that on a 12 mpx camera could be risky.
thank you very much for the reply. I’m sorry for the delay I have been out on site, I performed a Phantom 4 Pro flight this week and it had 1242 image and all in on the Time including spending around an hour to get the CGP’'s selected through the Images it took 7 hours to create everything.
So unfortunately I couldn’t start my next phase until 2 am which is okay every so often but I imagine I’m doing something wrong somewhere.
It did seem quite a bit quicker when I wasn’t using GCP’s I’m not sure if this adds to the time. I’m not sure that this is a particularly large model compared to other peoples models is this time of 7 hours about the same as others?
I have got the Minimal Distance between two Points at 0.002000 I wasn’t sure if I should have a larger increment at all.
The Phantom 4 Pro I’m using for this is a 20 Megapixel so maybe I can try and change the scale a little further
I am sorry to keep asking I just can’t really find many details about all of these settings.
Just wondering if any of these change the Accuracy down at all?
As long as the alignment is done well and the mesh does not look blobby or muddy the accuracy should not be changing, you are just decreasing the amount of detail. I cannot compare your speed to anything as there are thousands of factors varying. You can move the value of 0.002000 even further to 0.05 I think. It does need a bit of testing really. I cannot speak in general due to your framing thanks to height you shoot from is different your hardware both on site and in machine is different. I recommend you to start heavy on the settings and then move lower on them so you can get quick tests out and see whether you like what you see.