Face Recognition Based On Deep Learning In Natural Scene

Recently,with the development of artificial intelligence and the popularization of video monitor system,face recognition is of great significance in the security industry.Face recognition technology belongs to image recognition direction in the field of computer vision,is a biometric identification technology based on human facial feature.Compared with other biometric technologies,face recognition has the advantages of interactive friendliness and convenience.Although face recognition in constraint scene has reached the commercial level,but face recognition in natural scene has difficulties,due to the angle,lighting,resolution and other factors,thus,it is of great theoretical significance and practical value to search for a method of face recognition in natural scene.Firstly,this paper expounds the current research status of face recognition algorithm,introduces the principle of face recognition algorithm at home and abroad,and summarizes and analyzes the difficulties in face recognition technology.Secondly,this paper analyzes that the current face detection algorithm is time-consuming and limited by natural scene,and proposed two levels of human face localization algorithm.After testing,this algorithm can achieve real-time natural scene image processing and maintain a high face detection rate.Again,on the basis of face detection,this paper introduced the face automatic calibration algorithm and uses it to implement a face recognition based on multi-scale,high dimensional feature using Gabor wavelet.The experiment proved that this method works under the constrained scene,but does not work under natural scene.Finally,this paper summarizes the algorithm limitations of previous chapter and improves the DeepID1 face classification network to realize the multi-network model identification algorithm of the global face feature joint the local feature.It is proved that this algorithm is effective for natural scene.