Log in or register to post comments

bichromatic image (b/w) as trackable

February 23, 2013 - 5:30pm #11

Hello, I need to create a trackable that is bichromatic (two colors only), so I'm trying with white background and black shapes, a sort of reticular form, that must be interconnected because I must 3D print it.

I tried to use the stained glass photoshop filter with a photograph, and the resulting image, faceted multicolor, with a black irregular reticular, gives me a 4 starw rating.

So I monochromed the image, that is now the same black reticular over white background, (better contrast, should be better) gets 0 stars.

If I check for features, the bw image has even more features than the color one, but still 0 stars.

What's wrong? Why with so many detected features is 0 stars?

Can you suggest a way to create an excellent trackable with only 2 colors? (1 color over white background, but the design must be fully connected)

thanks!
 

 

bichromatic image (b/w) as trackable

February 24, 2013 - 7:41am #10

Reticulation can result in false positives in the target analysis, because the image will have many corners but their distribtution is too regular to be useful. If you can post or PM me your image, I'll give you some tips on how to make it both trackable and printable.

bichromatic image (b/w) as trackable

February 24, 2013 - 3:33pm #9

Hello David,

here you should be able to see the two images:

https://www.dropbox.com/s/nogpmybhbtv2j8j/stained_15_5_0.png

https://www.dropbox.com/s/t2rlkyt95th9clz/stained_bw_15_5_0.png

the colored one gives 4 star, while the same one bichromatic gives 0, and the reticulation is the same.
The bw seems even to have a higher number of features.

Moreover, I understand the distribution issue, but as you can see the color and bw have the same reticulation, why one gives 4 and the other 0 features?

thanks,

mattia

 

bichromatic image (b/w) as trackable

February 25, 2013 - 9:22am #8

That's interesting. Have you tested the B/W target?

I'm going to check on why there is such a disparity between the two.

bichromatic image (b/w) as trackable

February 25, 2013 - 9:35am #7

I think that the contrast variation in the color image is serving to break-up the regularity of the reticulated pattern. Take a look at the grayscale versions ( show features ).

color

Loading ...

 

b/w

Loading ...

 

bichromatic image (b/w) as trackable

February 25, 2013 - 2:48pm #6

I haven't tested the trackables yet, because first I wanted to understand if the evaluation of the target manager is correct.

I can't see your embedded pictures (it writes Loading... on both) but I guess are screenshots of the "Show features" button in the target manager. I tried it and it seems pretty clear to me that there are way more features in the bw than in the color one.
And the features present in the colored trackable are exactly the same of the ones in the bw (reticulation is the same, and is not so much regular).

I'm just trying to understand what determines a feature and what exactly determines a rating, because what I see is, for example:
colored: 100 features in a irregular reticulated pattern  4 stars
bw: the same previous 100 features in a irregular reticulated pattern, plus evan more features  0 stars

can you help giving an explanation? is the evaluation broken in this particular case?

tomorrow I'll try to print the bw and make a test, anyway

thanks

 

 

bichromatic image (b/w) as trackable

February 26, 2013 - 2:34pm #5

I tried the markers, and as the target manager says the colored 4stars one gets detectedmuch better than the bw 0 star one... as expected.

Still can't understand why since:
- the "show features" button actually show more features in the bw more contrasted picture
- reticulation is totally the same

so I'm unable to design an efficient bicromatic marker.
I also tried to insert in the target manager one of the frame marker included with vuforia, and still I get 0 stars...

 

bichromatic image (b/w) as trackable

February 27, 2013 - 4:36pm #4

If you look at the greyscale version of the colored target, you'll notice that the contrast variation helps to break up the regularity of the pattern. I think that's why it's performing better. The tricky thing about reticulated patters is that they will show a lot of features, but the regularity in the pattern of their distribution actually makes them difficult to track.

Does your target need to use that sort of pattern or is that a design choice?

bichromatic image (b/w) as trackable

February 28, 2013 - 1:23am #3

No I need to make a 3D print of the target, by printing a flat white square and a black shape with the actual target design.

When joined, the two parts make the marker, but the black one (just like the reticular) must have holes in it to show the white below, just like the bw marker I tested (I thought it was irregular enough).

The problem with the black shape is that is must me fully connected, so a sort of reticulation is a necessity, maybe I should try to make it even more irregular.

At the beginning I planned to use the frame marker instead of the image marker, but results are not so good (my marker won't be larger than 3/4 cm, and with a 4 star image marker I get very good results even with this size).

Can you suggest a way to make a bw only design that gets a good evaluation in the target manager?
What's strange is that other AR frameworks I tried before Vuforia used even a checkerboard as pattern, that's even more regular than my reticulation.

bichromatic image (b/w) as trackable

March 1, 2013 - 8:08am #2

PM me w/ your email and I can send you some examples.

bichromatic image (b/w) as trackable

March 1, 2013 - 10:21am #1

sent thanks!

Log in or register to post comments