02-22-2012 06:49 PM
Dear Sir,
My name is Chakkrit and I am a Ph.D. student working in the National Electronics and Computer Technology Center. I am currently using LabVIEW 2009 with NI LabVIEW Vision 2010 especially the Pattern Matching for my Ph.D. work. I understand that Pattern Matching in LabVIEW utilizes a normalized cross correlation which in theory gives a correlation signal between -1 to 1. Nevertheless, the correlation signal practically seems to be compared with a score value ranging from 0 to 1000. So, I would be greatly appreciated if you can give me more detail about the relationship between the score and the correlation signal.
Best Regards,
Chakkrit
02-23-2012 03:02 PM
Have never found an explanation of the details, but a score of 1000 means 100% identical. A score of 800 would be 80% identical, etc...
It would be nice if the algorithm details were much easier to find.
02-23-2012 08:51 PM
Thanks for the suggestion of Randall Pursley.
I tried to find the source for the reference of equation conversion from -1 to 1 ----> 0 to 1000.
For example: journal, conference, books, or expert communication etc.
Best Regards,
Chakkrit
02-24-2012 07:13 AM
Hi Chakkrit,
The mapping of correlation value to the score of 0 to 1000, which Pattern Matching gives as the output, is a scaling of correlation value of 0 to 1 to a score of 0 to 1000. If the correlation value is less than 0, it is assigned score 0. Please keep in mind that the Pattern Matching functionality is optimized over the area in which it performs correlation. So, it might not give exactly the same score as you get after correlating the entire template for a particular match.
Regards,
02-24-2012 09:58 AM - edited 02-24-2012 10:01 AM
Thanks for the great suggestions.
I tried to compare results of labview and theory, because it cannot sure that is correct which evidence used to authentication very important for me.
Best Regards,
Chakkrit