Hello,

I've put blobDetection code I developped last year ( http://v3ga.net/show.php?id=12&type=0 ) in a library.
I've slightly changed the interface and conventions. It is possible to have a custom method to accept/reject blobs, thus allowing to filter them according to your own criterias.
The library is documented and two examples available for download are provided.
Any thoughts or comments are welcome, hope this library will be useful for the community!

http://v3ga.net/processing/BlobDetection/Library/

Hmm, I have tried out your webcam demo and I get:

java:38:3:38:43: Semantic Error: No accessible method with signature "setBlobDimensionMin(int, int)" was found in type "blobDetection.BlobDetection".


Any thoughts?

So fast!  Thanks for putting that up.  I think many a folk will come to appreciate it over the coming weeks.  Certainly makes object tracking a breeze!

Hello Robert,

yes it is possible that the detection dies because of too many lines per blob (the blob number is limited to 1000 for each instance of blobDetection , ie when max numbers of blobs is encountered, the loop breaks, may be it would be good to let user set this parameter).
Each blob has a stack of 2000 lines allocated at the beginning, and the algo didn't check for overflow when running. I've quickly patched this , setting up the number to 4000 and checking for out of bounds. In this case, I think that it would lead to a 'truncated blob' as it stops. May be it would be good to have another better strategy there (ie allocate some more memory to let the detection finish for that blob).
Thanks for having pointed this out! I've uploaded a new version online :

http://v3ga.net/processing/BlobDetection/Library/jar/blobDetection.jar

Hope it will work as expected!

For loadPixels/updatePixels, there is no reason to let them there. I will update sourcecodes too on the tutorial page. Thanks for the tip!

Julien

so, what would one have to do to change the drawing of lines into the places of vertices within a beginShape(POLYGON).  I tried making the switch, but could not achieve the results I wanted... too many odd lines, though the effect was a bit puppetry-like.

r

I updated the blobDetection library. Now it is possible to retrieve polygon datas (ie a list of triangles) for each detected blob in addition to its bounding box and edges list. I used this new feature in two opengl applications. You can find more here :
http://v3ga.net/blog/?p=14

Library (documentation + download)
http://v3ga.net/processing/BlobDetection/Library/

Cheers.

Hello minga,
I didn't mentionned it in the blog post, but the video provided with the examples (Video & Kaleidoscope) were created using Quicktime 7. It may be the source of your troubles... Did you try loading another video?

Hi v3ga,

Some questions for you ;-):

1. Is that possible to set the minimun size of a blob. In order to not "compute" blob that are smaller than a defined size.

Something like:

static void setConstants(int blobMaxNb, int blobLinesMaxNb, int blobTrianglesMaxNb, int blobMinSize, int blobMaxSize)

2. In the doc you write:
"This library does not perform blob tracking, it only tries to find all blobs each frame it was fed with."

I'm not sure I get it. This library is doing blog tracking. no ? I mean with it I'm able to track a blob (in a real-time video stream), and therefore to know its coordonates, size.
What do you mean by blog tracking ? Tell me ;-)

3. Is the result I'm getting correct after doing the following initialization (see screenshot http://www.smallfly.com/v3ga_BlobDetection):

 theBlobDetection.setPosDiscrimination(true);
 theBlobDetection.setThreshold(0.9f);

shouldn't I detect only brigth blobs whose luminosity is superior to 0.9f (only white). And in this case (screenshot) only one blob and not two.

thanks,