Research On Face Detection And Recognition Based On IOS Platform

Face recognition is used as a way of identification and authentication by using biological features, it has a lot of advantages, such as safety, reliability and friendliness. At present, with the information security becoming more and more important, face recognition has become a research focus in this area. Furthermore, as the popularity of intelligent terminals and performance improvement, as well as the rising of mobile payment, the requirement of combing face recognition and mobile application increases greatly.Face detection is the precondition of face recognition. This paper focuses on the method of face detection based on Haar-like features, and introduces the Adaboost algorithm in detail. With the open source computer vision library of OpenCV, this paper realizes face detection system based on Haar-like features on the iOS platform. What’s more, according to official documents provided by Apple, this paper also achieves face detection system based on the framework of CoreImage. Through the face database of FERET, the face database of ORL and the self-building face database, this paper does a large number of experiments under the conditions of simple background and complex background, the results of experiments show that the face detection based on Haar-like features has a higher detection rate and it can be fit for complex environments.In the respect of face recognition, this paper mainly studies PCA algorithm, LDA algorithm and the algorithm of combining PCA with LDA, and realizes face recognition system based on three algorithms on the iOS platform. Through the face database of FERET and ORL, this paper does a lot of experiments for testing recognition effects. The experimental results can be summarized as the following three aspects:Firstly, face recognition system based on the LDA algorithm and the algorithm of combining PCA with LDA on the iOS platform has higher recognition accuracy than PCA algorithm in the face database of FERET and ORL. Secondly, comparing the recognition accuracy of three algorithms corresponding to two face databases can be found that recognition accuracy of face recognition experiments on ORL face database is relative low, it indicates that face samples included in the ORL database have a greater difference. Finally, the algorithm of combining PCA with LDA is superior to the other two algorithms in recognition accuracy and recognition time. What’s more, it has good robustness in the aspect of variations of facial expression, posture and illumination factor.