Implementation of an Efficient Access Control System for Secure Building

Abstract

This paper describes hybrid access control system based on a new approach of face recognition and PIN (Personal Identification Number) code identification, which is done by finding the closest match between features implemented on the same databases. This approach involves two identification and matching subsystems. The outputs from these subsystems are combined to decide the final output that authorizes the access to the secure building for certain persons while denying the access for the others.
In verification, a biometric identifier is compared to another template through a PIN number of database. Identification involves comparing an existing image against a large image database in order to identify the individual depicted. For this purpose, a camera is set in the door looking at the enter and the captured frames are analyzed. The implementation includes neural networks for PIN code identifications, and analysis starts by selecting the regions of interest in the face images using NN.
A neural network based upright frontal face recognition approach has been presented in this paper. The algorithm of face recognition is implemented using preprocessing of face image and two-stage neural network. The first net finds the eyes region of a person and improving image by histogram equalization. The second uses an image of the area around eyes region to identify the person. Using arbitration between multiple networks with rarely overlap in region images to clean up the results significantly improves the accuracy of the recognition. A feasibility investigation and evaluation for face recognition based solely on new concept is conducted which covers all conditions of human face recognition under varying lighting condition, varying facial expression, and varying pose.
The proposed system structure is implemented with newly proposed methods that allow fast processing and accurate recognition as a software package using C++ and Visual Basic.