Messages

1

General / Re: Photogrammetry With Moving Light Source

« on: April 21, 2016, 12:28:14 PM »

I've never worked with moving light sources but moving the camera relativ to the light source (e.g. the sun) should give you similar results I think. If you can influence it, I'd go for diffuse light and go ahead with the workflow. It also depends on the use case - what exactly are you trying to do?

2

General / Re: Merging Projects by GEO Coordinates

« on: April 21, 2016, 11:50:46 AM »

Hey there. You can actually merge dense clouds by "Workflow" -> "Merge Chunks" and set "Merge dense cloud" and "Merge markers" as active. I would duplicate the chunks and merge the duplicated ones so you don't lose data if it goes wrong.

3

General / Re: Good PS hardware setup?

« on: July 03, 2015, 01:48:23 PM »

I don't know if anandtech is anywhere near reliable but they benchmark various CPUs for every time consuming step on PhotoScan.

How many slots you have left for RAM? More is always better than less. I recently ran out of RAM while having almost 300GB. What exact GPU do you use? When it comes to CPU, I don't know how much to expect from a new and faster one. I would priorize GPU over CPU.

4

General / Re: How is the sparse cloud created?

« on: June 24, 2015, 11:18:57 PM »

Thanks a lot! The explanation is really easy to understand. Just as you said we probably are pretty far from "how things exactly happen" and since my time is really limited, I will maybe write some bits about bundle adjustment but not about e.g. epipolar geometry since I'd probably have to explain the mathematical background and that's a minimum of one more hour of work I can't afford at the moment. Also RANSAC and SIFT are really interesting from a neutral point of view but as I've mentioned, I don't have the time to explain all that. Regardless, I'm really thankful that you've explained things and helped me open up my horizont!

5

General / How is the sparse cloud created?

« on: June 21, 2015, 04:36:59 PM »

Hello fellow PhotoScan users,

I'm currently working on my bachelor thesis about creating historical 3D-models. While the work with PhotoScan itself was no problem at all, I have issues understandig "how" some things work.

As far as the users manual goes, PhotoScan is finding the right camera position and orientation for every single picture and builds a sparse cloud. This part is practically where the magic happens, right? How exactly do you get three dimensional points out of two dimensional pictures by determining the camera positions and orientation for the single pictures?

Thanks in advance. I'm really thankful for every answer and source I can get!

/Edit: Just to let you know - I've already used the search and only things I could find were these two topics (which still deal more with the algorithms itself than with how the 3D coordinates are found):
http://www.agisoft.com/forum/index.php?topic=89.0
and
http://www.agisoft.com/forum/index.php?topic=3713.msg19379#msg19379