Is it possible to merge point clouds to complete the model? Another tank scaning

[lextorlex]

Hi all. I'm knew with photoscan, but desided to test software scanning another tank (SU-100). I was able to make 200+ photos with Asus phone. It was difficult to get this photos because it was standing between other tanks and the weather was partly cloudy.  I loaded all of them in photoscan and failed to get a good result. only small part of the tank was recognized but even it was a quite wrong. Then I decided to devide project on parts and started to work with pictures which concentrated only on some parts of the tank, backplate for example. And you know I started to get quite good results. The problem is that now I have several photoscan files which reconstruct the whole tank by parts. I can export them all in 3ds max for example, then scale rotate and align by hand. But I wonder can point clouds from this files be merged and build final model. May be Photoscan can do it already? What I think about is smth like Artec Eva's sowtware align scaned parts to reconstruct the model. Any Ideas?

 

[lextorlex]

this is examples of part reconatruction.  All_cameras.jpg in previous post is the result of building model with all photos.

[igor73]

Yes, sometimes it helps align to divide up a project in chunks.  A tank seems like a good object for photogrammetry and quite easy to do if you do it right.    Think you should try with a good camera before you spend to much time with mobile phone images.   

There is indeed an  aligned feature in Agisoft as well as a merge function.  If you have no overlapping images between chunks you will have to use point based align, quite processing intensive but low res images should be fast regardless. 

[bigben]

Some of what you are seeing may be due to PS failing to make connections between images which can happen when you have a lot of them. These breaks can occur from large movements between images or low image quality in a number of images in the sequence. 

From the first model, you could find which cameras were used to make the model and then move all of the other images to a new chunk and align those

Alternatively, you could try adding control points to link those images to the others which you may have had to anyway for aligning multiple chunks.

Using calibrated lens settings may help in this situation as well.