Edge detection in Flash

Juni 30th, 2009 • ar, flash, math

here is a good post about edge detection with the sobel-operator

sobel-edge-detection-in-flash

Works really well, but does not fit my needs.

Ar without flash

Juni 28th, 2009 • 1 comment ar, creative, demo, hot

Flash sucks ;)

This here is awesome!

This is stunning – sorry no youtubed one available

platipus.avi

ar research

Juni 25th, 2009 • 1 comment ar

augmented realty in flash is propably one of the hottest topics for webdeveloper. you can implement it. BUT – does anyone of you know – how the math works?
well i don’t. but i will try to find out.

why?
hm – hard to explain. flartoolkit is great, but there are things in it i don’t like. it isn’t really good coded & you can only track that ugly black / white markers. maybe i can solve that. maybe not. but i will try

vertex detection in quad

November 1st, 2008 • 2 comments ar, flash, math

Problem:

You have a quad outline as point data, and want to detect the four vertices

I googled that problem a lot but couldn’t find any solution that fitted my needs. Moreover I didn’t want to copy paste something I do not really understand, so I solved it my way

Read more »

My ArToolkit

Oktober 30th, 2008 • 3 comments ar, flash, hot, lab

Short after midnight I started to do what i promised : Rewriting the arToolkit – in my way – for flash. Well now it’s 7am and I’m to tired to talk about details right now, but the major detecting work is done. The next big thing is converting the 2d data in 3d roll pitch yaw parameters (this will be pretty nasty math) and of course cleaning the whole coding mess up :D

When it’s all done I’ll post more details. Promised!

Update:

So whats better?

• works in bad light conditions
• needs less cpu performance
• the tracking is more stable (no jumping marker)

Whats up next?

• better edge detection
• find marker start orientation (black spot in white square)
• cleaning code
• coding the 2d 3d bridge

arToolkit and papervision3d

September 25th, 2008 • 8 comments ar, creative, flash, hot, papervision3d

Too sad that my french and japanese is really bad. I recently found a great arToolkit demo for flash called FlarToolkit. After randomly clicking all the links I could find on the page, I found the svn :-)

• checkout the sources here
• blogpost here

I had a compile error using flex sdk and eclipse. after changing a private to public property it worked.

It performes really well but you can also see that I is still not really stable. I think the major bottleneck is the bad webcam quality compared to a dv cam.

What I don’t get:

When the FlarToolkit has found the rectangle in the “2d” image, how can I transform this data in the papervision3D room. If you have a 2d->3d math for dummies link – please drop me a line.

What I would love to do:

As you can see there are some serious tracking precision issues. In my opinion the major problem is the pixelbased tracking. It’s a waste of performance. This could be solved by an Idea Mario Klingemann presened on the FFK08 called blobdetection.

Well this is my theory:

1. Blur the camera image and convert it in black / white image
2. With getcolorBoundsRect and floodfill detect all black rectangles in the image
3. Find the marker rectangle
4. Use an edge detection
5. You should be done

So you only need to analyse a small amount of pixels directly for the edge detection. The other stuff can be done by “internal” flash functions which are really fast. Maybe I’ll try it – someday.