Research On Troops Roll Call System Based On Face Recognition

In today's society,with the development of information technology and the improvement of human's safety consciousness,the information security and concealment have gained considerable attention.Face recognition as one kind of biometric identification technology is reliant on people's facial features information.Comparing with the other biometric identification technology,it is more convenient,rapid,noninvasive and interactive.Moreover,it has strong ability of tracking.Therefore it's widely studied and actually used in the field of finance,military,government,aerospace,public security,etc.As the armed police squadron at the grassroots level,superior has to regularly count members.This paper implements the troops roll call system which was based on face recognition algorithms.Moreover,we proposes a novel weighted and blocked LBP feature,which can in some extent to improve the accuracy of face recognition in the experimental results.Specifically,In terms of face detection,the current popular Haar-like as the mainly feature for face detection.Meanwhile,we adopt AdaBoost classifier to choose the feature which can be most representative.Besides,histogram equalization and face correction also are adopted for improve the speed and increase the accuracy of recognition process.In terms of face recognition,the approach of weighted blocked local binary model is introduced as the mainly feature.First of all,after face detection and correction,each face image was partitioned evenly,and then extract LBP features from each image block,then use a classifier to training and predicting for recognition rate.After that we calculate weights for each of block LBP feature.According to the high or low recognition rate,we give the corresponding weights.Generally speaking,the higher recognition rate is,the larger contribution to the facial expression are,then we would like to give a greater weight,on the other hand,we give a less weight.By such a mechanism,we can maximize the distinguishable characteristics of face image.Finally we concatenate each weighted feature from each block to form the final face feature presentation.Having done all of these,use another unrelated dataset to test the effectiveness of our method.We through a lot of experiments to verify our method and we also establishes multi-face of recognition system in the mobile terminal.Mobile side face recognition system can not only meet the requirement of real-time used in roll force,but also has several benefits,such as,convenient,easy to operate,etc.