METHOD AND SYSTEM FOR RECOGNISING FACES BASED ON LIST OF PEOPLE NOT SUBJECT TO VERIFICATION

摘要

FIELD: physics; signaling.;SUBSTANCE: invention relates to biometric systems for personal identification from a facial image. In the method a collection of images of "negative" people - those people who are not subject to verification is created. A collection is created for known persons with conversion of the facial image into an internal presentation applied for recognition algorithms realised in the method. The face of a person is detected on an input image. A segmented region containing the face is prepared for recognition through standardisation of the position of the face in accordance with the position of eyes. Scaling to given dimensions is done, as well as standardisation of intensity. Fundamental face recognition algorithms which can calculate similarities between two images are executed. A subset of "negative" faces is selected and then used to determine a new recognition algorithm. A new recognition algorithm is created based on conversion of degree of proximity of the fundamental algorithm based on the created collection of "negative" faces. Recorded images are classified into a set. The system realises the said method.;EFFECT: personal identification in non-uniform illumination conditions and for a wide range of angular displacement of faces on an image.;2 cl, 4 dwg