According to the single should be between the two images H matrix and calculation of two feature points is repeated, set the correspondence variable here, as the number of repeat feature points and the matching results evaluation of two parameters Precision and Recall has the following formula:


Formula, # correct matches means to match result in correct matching point to, # correspondences said repeatedly while feature point detection feature points to (repeat feature points that the actual for the correct matching point, but may be matching algorithm, also may not match), # false matches said wrong results in the match point to, among them, the right and wrong matching point to can use the above mentioned two conditions to determine whether feature points repeat,

Therefore, when evaluating feature point matching results, first of all should know should single matrix between the two images H, and then through the feature point detection algorithm to get the feature point information about image, according to the characteristics of single sex matrix should repeat points, namely # correspondences; Then a feature points matching algorithm to get the matching calculation Precision and Recall as a result, draw 1 - Precision and Recall curve, the curve on the top of the results are good!


Here is how to find the correspondence, and what is the right match? (what kind of match is # correct matches),
Have a great god wrote the code related, can share with you!

CodePudding user response:

The original poster hello, do you have solved your problem?