Multiple Human Face Detection and Recognition based on LBPH and Machine Learning

Abstract

Face recognition failure is one of the challenging hurdles that a machine deals with. As the revolution of technology comes into the picture, the usability of digital security based on facial biometric become most frequent today. The obsolete security system based on the recognition of features is unable to tackle the variations in a dataset. The failure rate in the face recognition system is found to be common in the various digital systems. This paper introduces a robust design to tackle the recognition issues. This paper utilizes a series of robust techniques in order to establish accurate face recognition with the minimum failure rate. In the proposed design, the facial portion of the acquired image is segmented out using the Haar cascading algorithm which deals with the pixel values of the image. The extracted facial portion of the image is normalized using the normalized pixel intensity algorithm to minimize the noise ratio. Then, the feature extraction procedure is applied to the normalized segmented facial portion of an image which is done by local binary patterns histogram algorithm (LBPH). The face recognition is acquired through such extracted features using the AdaBoost algorithm. The proposed design of face recognition is explored against the testing image with the random variations to generate the accuracy which is found to be 96% and it is proven better than other existing algorithms.