Face recognition has many applications: Forensic science Security systems Border control Driver’s license and passport verification Access control Localization of missing people Our algorithm is based on the Analytic Hierarchy Process. The process is used to pairwise compare two facial images/facial features/parts of the face. Imagine that you say that the eyes are much more important that the mouth in the process of recognition. Then you can write, e.g. eyes:mouth 7:1. This way we can get the matrix of pairwise comparisons. In each row and each column of the matrix you can input the result of comparison. This matrix is called “reciprocal”. Finding the maximal eigenvalue of the matrix and corresponding eigenvector is the most important thing here. The normalized values of the eigenvector’s elements are the weights corresponding to the rows/columns of the matrices (and to the compared features). Using these weights you can build the fuzzy sets and the membership functions which are the representatives of the fuzzy sets. If we want to find someone (who, for instance, we don’t know) in the dataset we have to compare her/his face image with other images in the dataset. To compare two images you can just compare the values of the membership functions. Watch my interview at YouTube :) youtu.be/M3RN1GkccjE
Posted on: Tue, 18 Nov 2014 09:03:15 +0000
Trending Topics
Recently Viewed Topics
© 2015