wrong lake detection

[stamatis]

HI. my name is stamatios and i am kinda new at the forum.I choose IRIS recognition for my final year project and i would like to ask you some things.

I am using NI VISION ASSISTANT to measure and print out some results.I managed to design a detection in the eye that measures and printing out some results.As u will see in the word document that i have in the ZIP file i wanted to measure the lakes in the eye but in my desing it measures lakes but not the proper ones.Did i use wrong filters or do i have to change something?If u can plz show me the correct filters or somehow to tell me whats the wrong with this design.Another problem is when i load another image (eye) with eyelashes and apply for threshold it measures the eyelashes as well.Is there another way to remove the eyelashes or to measure the eye without measure them?

Than a lot stamatios

[Robert_Lee_NIUK]

Hi Stamatis,

Thanks for posting on the NI forums. I have had a look at your word document and while I can quite clearly understand what you are trying to do, we need to focus more on the specific code you are using. The other file included (with .scr extension) cannot be opened so I am wondering what this is meant to be?? If you could perhaps post your code on the discussion forums then it might be a lot easier to see exactly how your code is performing the relevant filtering. With regards to the query on the eyelashes, the best method would be to remove the top section of your image and then perform the threshold readings.

If you could let me know exactly what NI Vision software you are using, that would also be of use.

[Robert_Lee_NIUK]

Stamatis,

My apologies, I didn't realise the file was a Vision Assistant Script file, I will look into this for you and get back in due course.

Best wishes

[stamatis]

Thank you very much for your reply.

[stamatis]

My problem occurs when I try to apply some filters using the vision software (v.7) in order to separate the lakes from the rest information. I havent managed to isolate the lakes so far, as you can see there is unwanted information as well at the final stage of the process. Lakes are unique in every person eyes as they are different in size and number. The lakes also reveal a pathological problem that ppl might have "iridology". that method is not 100 % accurate.

BTW how can I edit my posts?

cheers

[chilly_charly]

I just had a look to your problem. I must confess that I don't understand how you managed to recognize the "lakes" on the open first document. Unless you can explain clearly what a lake is and how you can identify it manually, I'm rather pessimistic about the issue of your project !

[stamatis]

The correct term for "lakes" is pigment spots and i havent managed to isolate them..

[chilly_charly]

---

stamatis a écrit:
The correct term for "lakes" is pigment spots and i havent managed to isolate them..

---

I can't see any specific pigment spots at the places where you draw the circles in your Read It first document. Seems that a strong expertise is required to guess anomalies. No wonder the technique is not 100 % accurate. I sincerely doubt you will be able to succeed using image analysis.

Message Edité par chilly charly le 10-28-2007 12:49 AM

[stamatis]

HI ALL,

Afte several tests i made on  my first Vision Assistant program i managed to implement it and make it more auspicious and operational.I created a LABview VI and i tested it but i still got some unwanted results like:

1:In my front panel the image(eye) it doesnt appear to the Image window properly.I need to move my cursor to x,y axes to find  it.

2:When i find the image i havent got the size where i want to.The pallete it has only zoom+ and not zoom-

3:When i get the image to the LABview window it doesnt appear the last step from my vision assistant where is the pattern matching.I want to get the last image from the vision assistant that shows the patterns..

Here is my  Vision Assistant design , the LABview design and the patterns...Can someone explain me whats going on?

[chilly_charly]

You can't expect to get results with such a method !!!

What you are doing is :

1/ The operator identify the lakes and define "templates" on a given image

2/ The machine search the same image for the lakes identified by the operator

Of course the second step is completely useless. This approach will never ever work on real images.

Moreover, in a previous post, you told us that the lakes were pigment spots, and here you are working with a grey level image ??????

I repeat, if you are not able to explain clearly how you can recognize visually a "lake", you are loosing your time ! Try to find an easier final year project.

Message Edité par chilly charly le 10-31-2007 02:22 AM