What is the accuracy of such reconstruction methods with regards to the vibrations of the flying drones, quality of camera and resolution? Is it possible to improve the results by organizing multiple flights and overlaying/accumulating the data in the point cloud? Is there any free software available?
Hi Reiner,
I would say the best method would be Structure-from-Motion (SFM) based methods. Ofcourse, the challenges with flying platforms would be vibrations and video stability but techniques exist to deal with these to great extent. Visual SFM by default would generate relatively sparse point clouds but with surface fitting and strong priors based reconstruction you can improve the results greatly, atleast in my experience. I would recommend VisualSFM (http://ccwu.me/vsfm/) to be a good starting point to experiment with.
Tauseef
Hi Reiner,
I would say the best method would be Structure-from-Motion (SFM) based methods. Ofcourse, the challenges with flying platforms would be vibrations and video stability but techniques exist to deal with these to great extent. Visual SFM by default would generate relatively sparse point clouds but with surface fitting and strong priors based reconstruction you can improve the results greatly, atleast in my experience. I would recommend VisualSFM (http://ccwu.me/vsfm/) to be a good starting point to experiment with.
Tauseef
Hi Reiner,
From my point if view Image Based Modeling is good method
Jayant
For an off the shelf package Agisoft PhotoScan has now gained enormous credibility with professional productions. I am not associated with them but I know many productions that have used it http://www.agisoft.ru/products/photoscan
What about Kinect or structure sensor? But I of course agree that this method are expensive in term of CPU and weight. However work only in a range between 2 to 8 meters.
Hi Reiner,
I would as well recommend SFM. The result will depend on your camera (e.g. resolution, sensor) and the camera network (e.g. overlap, camera position). I am using Agisoft Photoscan now but have used open-source packages such as VisualSFM. The advantage with Photoscan and other commercial packages is that you have a all-in-one solution (image processing, georeferencing, DTM generation etc.).
HTH
Klaus
Agisofr Photoscan (for me is the one... ask me for a free trial license)
Trimble UAVmaster
Pix4D
Ensomosaic
MenciSoftware APS
Photomodeler
Autodesk123DCatch
The two that would be at the top of my list is Autodesk Recap (it is essentially a pro version of 123d catch) and Agisoft. I use both of these packages for this purpose.
I am using Agisoft for model creation, so this would be my first choice for this.
VisualSFM is a little complicated to use at first but you can get some pretty impressive results from it.
for processing and analysing point clouds i recommend cloud compare.
http://www.danielgm.net/cc/
@Eik:
I am curious how good does you point cloud look like before you do the surface reconstruction? It would be wonderful if you could post a picture of the pc and surface reconstruction! And what do you do with noisy clouds?
Thanks
@konstantin
Here are some fast examples of Point cloud and surface reconstruction. It's made from 7 Pictures, the Point cloud has 1.3mio points, the mesh has 1.4 mio faces
You can see in the nearest neighbor pic (blue) that its extremly dense.
you can get rid of noisy clouds with some statistical tools from Cloudcompare, like FLATTENING or Removing Outliners. Also you can keep the noise down by choosing an aggressive depth filter setting (in Agisoft)
The best way is to avoid taking pictures that lead to noisy clouds. For this keep the ISO at a minimum level, do not use blurry Pictures....
Hope this helps
You could also used 123Dcatch from Autodesk (http://www.123dapp.com/catch) for reconstruct volume from pictures. It is a free software and could be a good alternative to perform some test before invest in a more expensive program.
Hello Geoffrey,
Autodesk 123d Catch has some promising results (how many pictures are possible? 50?). Still, I am always interested in the mesh itself and 123d Catch suggests a nice reconstruction with textures, but the mesh itself is often in need for further correction with additional software.
If you want to go one step further, I suggest using the - for now free - beta software of Autodesk called "Project Memento" (https://beta.autodesk.com/callout/?callid=D69655022B4F42C880513D4E3085C7D0) . It is said to use a better algorithm, no reduction of photo quality and an higher limit for pictures. It is unlike 123d Catch not a mobile application. Autodesk calls it a buisness all-in-one software (just an example: like 123d Catch with Geomagic abilities).
Still, there are a lot of softwares circulating with pros and cons and only few or more outdated publications.
Sincerely
https://beta.autodesk.com/callout/?callid=D69655022B4F42C880513D4E3085C7D0
iWitness PRO has the capacity to to simplify the point cloud which may help you but it does look like Agisoft has the most promise for large aerial scans. We have had good results with Autodesk Recap and 123D catch but have only used them for ground based photo's.
What about MIC MAC? Quite tricky to master but much more precise to my point of view. And, yes, it's free!
http://logiciels.ign.fr/?-Micmac,3-
I reviewed different reconstruction algorithms in the following publication:
Conference Paper A REVIEW OF RECENT RANGE IMAGE RECONSTRUCTION ALGORITHMS
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Computer Graphics and Animation