Comparison of normalization techniques for lightning independent face recognition

Face recognition has become a popular ring of the security industry. Identity verification in the banking sector has become a security measure used to detect, detect and generate alarms for many other inhabitants who require high security at airports, airports and similar areas. Therefore, face recognition problem in image processing has become a field of work by many people. One of the issues faced by researchers working in this field is the problem of illumination, ie face recognition weakness that occurs in different light conditions or in different exposure situations. The aim of this thesis is to solve the problem of illumination caused by various reasons on the face recognition systems designed for this thesis and to solve the failure and weakness of face recognition due to illumination by using various lighting normalization techniques. Using PCA (Principal Component Analysis), the success rates on the facial database of face images taken at different exposure angles and at different lighting conditions were calculated using face recognition. In order to obtain success rates, different parameter filtering methods of various lighting techniques are used. Different methods of the Cascade Vision Detector algorithm have been used in determining the face. The face database on which the tests are made includes right-lit, center-lit, left-lit, eye-free, glasses, normal, blinking face images. Facial recognition, one of the pre-processing techniques, was performed and the face was cropped from larger images. Illumination normalization has been shown to increase facial recognition success rate by approximately 55% in tested face databases. Keywords: Object Recognition, Face Detection, Face Recognition, Illumination Normalization Techniques, Filters, Information Security, Image Processing