Reconhecimento facial com os algoritmos eigenfaces e fisherfaces

Abstract

The objective of this work is to perform an investigation about the behavior of facial recognition algorithms described in the literature. We used two of the main algorithms described in the literature and proposed a set of tests in order to evaluate the behavior of these algorithms in order to obtain results that are compatible with a possible application of access control to physical spaces. The techniques used were the PCA and LDA, which are the basis of the operation of the Eigenfaces and Fisherfaces algorithms, respectively. For the trainings and tests of the algorithms, 3 different databases were used, to mention, YALE Database A, ORL (also known as AT \ & T) and Merge (made up of the junction of the two previous banks). For the analysis of the results, confusion matrices were constructed that aided in the evaluation of the algorithms against the number of false positives, critical data for access control tools. Of the two algorithms studied, the one that presented the best results was the Fisherfaces, which presented a better hit rate and with a minimum of false positives.