detección de figuras descriptores de Fourier

[rgarcíafernánde]

need help detection figures Fourier descriptors is for a science project

[StuartBruff]

Ramón García Fernández wrote:

need help detection figures Fourier descriptors is for a science project

I think you need to give us a bit more information than that! 🙂

What are you trying to do and how do you intend to use the Fourier transforms?  Which version of Mathcad are you using?

Stuart

[rgarcíafernánde1]

I need to recognize the type of figure geometric for the number of sides, perimeter and area by discrete Fourier transform one-dimensional

[StuartBruff]

Ramón García Fernández wrote:

I need to recognize the type of figure geometric for the number of sides, perimeter and area by discrete Fourier transform one-dimensional

Interesting.  Do you have a particular range of geometric figures in mind (I'm guessing regular polygons, but maybe you want irregular polygons as well).  How do intend to express the data for each figure (eg, radius versus angle about some point internal to the polygon, length of each side and associated turning angle, array of (x,y) vertices)?  Do have a model for doing the analysis or are you looking for ideas on that front as well as how to implement it in Mathcad?  What version of Mathcad are you using (it may make a difference if you're using Prime rather than, say, Mathcad 14 or 15).

Stuart

[rgarcíafernánde1]

Can I have your email to send my project?

[StuartBruff]

See the message in your PTC Community Inbox.

Please note that it's generally better to post your information here on the Mathcad Community (you can select the "Use advanced editor" option to bring up another webpage (dialog) that will let you attach Mathcad worksheets or other documents to a message).  You will get a wider range of people able to contribute and hence, generally, a number of different ways of tackling any given problem.

Please also note that if a message concerns homework or college course work, people will offer guidance and look at specific problems, but they will expect that the person posting will do most of the work (especially the typing up bit!).  Of course, if your problem is sufficiently interesting, then you may find several people saying "Ooh! Shiny!" and doing more than they should ... great for getting A+ marks, but not so good for learning by doing!

Stuart

[StuartBruff]

StuartBruff wrote:

...  Of course, if your problem is sufficiently interesting, then you may find several people saying "Ooh! Shiny!" and doing more than they should ...

Told you so.  Ooh. Shiny!

I just a had a quick look at one way of appying a one-dimensional measure to a polygon and then took the FFT.   It's top of the head, just to get a feel for the problem.  I haven't looked at any references to see how to classify shapes based upon the parameters you mentioned.   However, for regular polygones, the Fourier Transform shows a fairly obvious pattern. Nothing immediately leaps out with the irregular polygon, though. I suppose one of the next steps would be to add some jitter (noise) to the vertices to reduce the regularity.   I look forward to seeing what you actually want to do.

Stuart

[Fred_Kohlhepp]

A short Google search on Fourier descriptors led to a pdf paper from Microsoft--describing points on the boundary (x,) as x + y I, building a vector of points spaced along the boundary, then basically taking the FFT of that vector.  They were actually looking for the ability to recognize the same shape scaled, shifted, and rotated.  At that point it got too complicated and I quit, but it looked to be on point here.

[rgarcíafernánde1]

The objective of this project is to recognize a certain figure from the number of sides or area or perimeter which is to realize a new level pass That would recognize faces of people then , just so we must understand things like what are these for Fourier descriptors , methods that lead us to understand a little more what computer vision for pattern recognition

[rgarcíafernánde1]

https://mega.co.nz/#!5F4jnI6a!mNms-qhIgXlPrNqxCDcDxYhGje8HEsKzmmlp1h0NNCY

https://mega.co.nz/#!JV5ElabA!lQxBYnKYTH2nxYcaQ1g6XZamC_laE5NfrurASS_GA7Q

Two projects in which we have been testing

[rgarcíafernánde1]

https://mega.co.nz/#!FIJnDIzS!C05g3wstwxqdtbp73QtNlNehZqqKvqVgKRgAs33JQCY

[StuartBruff]

Hi Ramon,

It'll take me some time to sift through information that Fred's pointed me to and your worksheets.   However, after a quick browse through them, I was wondering if you might be better off looking at the built-in image processing functions that come with Mathcad 15 ... look under Help, eBooks, Image Processing.   This includes, for example, a number of edge detectors, including Roberts.   If you haven't got M15 or the Image Processing eBook, then you could speed things up a little bit by carrying out FFT convolution using the various kernels, rather than iterating through the image matrix. 

Do you have a particular reference that you are working from?   How exactly are you classifying and recognizing the images?

Stuart

[ValeryOchkov]

May by this on Mathcad Server

http://twt.mpei.ac.ru/MCS/Worksheets/Fourier.xmcd