OpenCV FaceRecognition prediction confidence values

edit

I'm working with a FaceRecognition algorithm using LBPH, and my code works perfectly (using Python).

What is bothering me is the prediction values. Through my code I get the confidence value, and I put that value on the screen. Usually that number varies from 30-75 from known people, and 120 -180 from unknown people.

I know that a high value means that the input image and the training image are not the same person. And values clos to 0 are better.

My question is: ¿How is calculed the prediction confidence value? (the distance between the images)

Is it just euclidean distance? or another method?

PD: Sorry for my bad english. PD2: I know there are some similar question but none answer which method is used to calculate the distance between the images.