Deep learning based secure face recognition with 3d depth camera

Deep learning and facial recognition techniques are a research subject that has developed very rapidly in recent years and has applications in many areas in daily life. Facial recognition systems can be used in areas such as tracking criminals, tracking personnel entry and exit from the company. Another important point along with face recognition is the prevention of attacks against face recognition systems. For example, face recognition systems can be misled by methods such as passport photos, printing a person's face photo from the printer, using a face photo from a phone or tablet, video images, and masks. Therefore, preventing the face recognition system from being deceived is as important as developing a successful face recognition system. In this study, a highly successful face recognition system has been developed using deep learning techniques. A system has been developed that provides security against attacks that are made to mislead face recognition systems by analyzing depth information and detecting blinking with a 3D depth camera. By calculating the gradient of the depth information obtained from the 3D depth camera, amplitude and angle histograms were extracted, and the vitality of the people in front of the camera was determined by performing statistical analysis of these histograms such as mean, median and standard deviation. In addition to depth information, blink detection was performed using the decision tree regression technique and the vitality detection performance of the system was increased.