AI ?

[Mohammad Dayyan]

Hi. Do you know any algorithms for detecting a shape in an image ? thanks in advance.

[Lost User]

hough transform : for curve responses harris coner and sobel edge detectors : for macro or constellation features this kind of problem is studied within computer vision.

[jk chan]

Hough transform can help you in simple shape detection like line circle etc.. You better look PDM(Probability density models ), these can detect shapes. PDM need to be initialized with a training set , after that it will calculate PCA . PDM is not that difficult.. have a look.

If u can Dream... U can do it

[Alan Balkany]

Mathematical morphology provides simpler and more efficient tools for detecting shapes and more. The fundamental operations are dilation and erosion. This http://www.ph.tn.tudelft.nl/Courses/FIP/noframes/fip-Morpholo.html[^] covers some of the basics (but it makes the concepts look harder than they are). This one http://www.dca.fee.unicamp.br/dipcourse/html-dip/c9/s4/front-page.html[^] has some nice pictures, but is short on explanation. If you could describe the task you're trying to accomplish, maybe I could make some more specific recommendations.

[Mohammad Dayyan]

Great, thanks

Alan Balkany wrote:

If you could describe the task you're trying to accomplish, maybe I could make some more specific recommendations.

Actually , I'm going to detect some shapes (likes circle or rectangles or complex shapes ) in an AVI file or a stream video from WebCam with C# . Do you have any advices ?

[Alan Balkany]

Once you get your image into a bitmap, morphology provides operations you can use to simplify the image, so it's easier to detect the shape you're looking for. For example, to detect circles with diameter d, doing an erosion by a diameter d circle should result in a single point for every circle in the image. Many systems have morphological operations built in, which is the easiest way to do it. You can also implement your own, but this is harder and requires specialized knowledge. I found a better link for morphology: http://en.wikipedia.org/wiki/Morphological_image_processing[^]. This includes links to free function libraries that do this.

[Mohammad Dayyan]

Thanks a lot Alan. What about http://www.aforgenet.com/[^] ? Is it a good stuff to do that ?

[Alan Balkany]

Yes, that looks excellent. It has morphology and much more.

[Mohammad Dayyan]

Thanks a lot Alan Balkany for answers.

[Alan Balkany]

You're welcome!

[PredictorX]

Assuming that these are simple geometric shapes on relatively simple backgrounds (as opposed to real objects against cluttered backgrounds), then I'd suggest: 1. Determine which pixels are foreground (shapes) and background (everything else) 2. Identify individual blobs (connected sets of foreground pixels of the same type) 3. For each blob, extract meaningful features 4. Train a classifier, based on the extracted features 5. Test the system on new images 6. Celebrate! In step 2, try using a flood fill. In step 3, there are many features one might try, such as perimeter-to-area ratio. In step 4, the classifier could be any learning algorithm: neural network, linear discriminant, etc. -Will Dwinnell Data Mining in MATLAB

[Mohammad Dayyan]

Thank a lot for the reply. I'm gonna using C# .