Home > Software engineering >  About the problem of calculation of matching rate in image feature matching
About the problem of calculation of matching rate in image feature matching

Time:11-18

Correct match and mismatch on is how to determine? Saw some papers, there is no detail,
Such as: a total of 100 pairs before use RANSAC algorithm, and then use RANSAC algorithm in 90, is the total matching refers to the 100, or 90?
If it is 100, then the correct matching rate is 90/100=0.9?
If it is 90, so is to manually from the $90 to find matching logarithmic error?
PS: because use RANSAC algorithm after the rest of the points is still very much (1000), and more intense, if artificial contrast can be cumbersome, at this point should be how to do?

Wish you a great god give some advice, I appreciate!!!!!!
Wish you a great god give some advice, I appreciate!!!!!!
Wish you a great god give some advice, I appreciate!!!!!!

CodePudding user response:

Your top, you dear friends to help

CodePudding user response:

I don't understand this

CodePudding user response:

I don't understand this, I say I used to practice, we used to be using sift feature matching surf, usually some feature points to run far away, textures and more complicated place and easy running deviation, I at that time, it is the image block, write their own algorithm into scratchable latex or 16 block even twenty-five dollars so, then do then sift in the area of the block, so, even if the running deviation, range will be smaller (image is more uniform, of course, this requires matching, heavy sampling and pretreatment and will do), the second is the characteristic points of smoke dilute, screening feature points is relatively evenly distributed

CodePudding user response:

As for the matching rate of the problem should be in the field of statistics, I think can calculate by 90%
  • Related