Log in or register to post comments

Setting up high density multitargets

March 24, 2017 - 5:01pm #1

Hello all,

 

So I'm hoping I can get a little bit of help finishing my project. I need some immediate assistance.

 

I've pasted 550 posters in a huge interior. I need to place all the image targets in the 3D scene to build the multitarget to create the SLAM using Vuforia. What I need is a faster way of "placing" all these targets in the scene.

 

I considered loading each plane as a single image target, and then viewing around the room using the camera. I would then copy and paste the positioning data of each image target by opening & closing the scene, (and saving the info).

I also 3d scanned the interior using an array of xbox one kinect scans, however I have to rebuild the scene. Also, kinect scans are notoriously bad for their accuracy with interiors

My question is; is there a faster way to do this? I have about 3000ft^2 that I need to convert, and I want to finish it in a few hours. The method I outlined would take hours, (if not days). I'd appreciate any help with this.

 

Thanks!

 

Setting up high density multitargets

March 28, 2017 - 7:31pm #3

I really need someone to give me an answer about this.

 

Basically I need to create a custom multitarget. I've already arranged image targets three dimensionally, and I have over 130 targets i need to combine to a single multitarget.

 

here's a better question; HOW DO I CREATE A CUSTOM MULTITARGET FROM PLACED IMAGE TARGETS IN UNITY. 

 

Instead of saying "edit the file", i need a link to a video tutorial, or a detailed example. all the other posts go to broken links. I find it very odd that multitargets make up a quarter of the functionality of this program, and there aren't any tutorials on it!

Setting up high density multitargets

March 24, 2017 - 8:26pm #2

I've read other threads asking similar questions. Basically, this is a question about adding a multitarget with more than just 6 faces.

 

From what others have said, you have to manually edit the XML file with the rotation and position of each image target. There are two problems to this method:

 

1- There is no information in the XML file. When I download the database and view the xml file, I get this:



<ImageTarget size="8.500000 11.001963" name="IMG_0535"/>

<ImageTarget size="8.500000 11.001963" name="IMG_0534"/>

<ImageTarget size="8.500000 11.001963" name="IMG_0533"/>

<ImageTarget size="8.500000 11.001963" name="IMG_0532"/>

<ImageTarget size="8.500000 11.001963" name="IMG_0531"/>

<ImageTarget size="8.500000 11.001963" name="IMG_0530"/>

<ImageTarget size="8.500000 11.001963" name="IMG_0529"/>

<ImageTarget size="8.500000 11.001963" name="IMG_0528"/>

<ImageTarget size="8.500000 11.001963" name="IMG_0527"/>

<ImageTarget size="8.500000 11.001963" name="IMG_0526"/>

 

As you can see, there is no place for me to edit the XYZ+rot information in the data file. I'll upload it so you can verify yourselves.

 

Second- Even WITH the ability to edit these coordinates, it's still tedious to modify large data sets. How do I get the initial positions of the image targets? I figure I could do this by manually creating image targets for the 500+ image, and then running the scene. During the scene, I would copy the coordinates and scale factor of the various images about 10 at a time, and then repeat the process until I figure out the approximate position of all 500. doing so would take hours, (if not days).

 

There are additional problems which I face to get this done. To begin, doing this with over 500 images will take about 3 hours of work to create the image targets in unity ALONE. The software is very poorly written in unity. Large datasets force you to scroll one possible solution at a time. This is ok for the first 50 images, however once you get up to 200, it takes you a good 2 minutes just to select the correct file. There needs to be an automated way to drag+drop converted the image targets into the scene. 

 

In my experience, the only way to achieve this project will be to use my kinect + kscan3d to build single frames, and manually align each reference mesh. This is the method I used for the initial prototype I did in my garage. It took hours to do, and that was just 10 targets! I'm hoping there's a much faster solution to building an interior 3d space. Furthermore, instead of using multitarget, I instead parented 5 image targets to one-another, and only had 2 active at a time. I have no idea how to get multitarget to work properly, let alone with hundreds of potential targets.



If someone could please assist me on this immediately, I would greatly appreciate it. I only have 2 more days to get this done, and I need ways to expedite the process.

AttachmentSize
Package icon Holodeck_1.zip1.75 KB
Log in or register to post comments