Binary Pattern Descriptor Based Face Recognizer for Access Control System with Real-time Dataset Generation

State-of-the-art biometric based access control systems like fingerprint and retina scan systems are becoming ubiquitous. In this novel-work, we propose to use a common camera to generate user-data in real-time and store it on cloud for access control of the user to any premise by face recognizer working on the principles of Computer vision, Image Processing and Artificial Intelligence. It may seem tedious and challenging to gather the large set of data which could train our model for better accuracy but in this work, we have generated the training data in an efficient way by using the concept of real-time data set generation. We have used the frontal face Haar-Cascade classifier to detect the object for which it has been trained for, from the source. The Haar-Cascade is trained by superimposing the positive image over a set of negative images. For training we have used Local Binary Pattern (LBP) algorithm which is a simple yet very efficient texture operator which labels the pixels of an image by thresholding the neighbourhood of each pixel and saves the result as a binary number in a CSV file. Once we have the dataset in structured (CSV format), we apply the AI algorithms to detect the current face by naming them (multiclass classification). If some unknown person (no training data available) happens to be in the frame, he is labelled as 'unknown'. Thus, a trigger is created by comparing names of objects and names in our data set to grant access.