| Face is a common and complex visual model,the human face reflects the visual information in the exchange and exchanges between people has an important role and significance.The processing and analysis of human faces have a wide application prospect in video surveillance,import and export control,video conferencing and human-computer interaction.At present,the main directions of face image processing include face detection and face recognition.Among them,detection as a special case of object detection problem has long been concerned.Face detection is the key step in face image processing,so the accuracy of face detection algorithm directly affects the follow-up work.The core problem of face recognition technology research is to make the computer have the ability of identification,face recognition is based on the results of face detection,carried out a deeper level of image understanding technology.this paper studies the face detection and recognition based on the static face images.The face detection algorithm is based on Adaboost algorithm,and its training method and detection process are improved.The multi-resolution idea is combined into Adaboost,A multi-resolution Adaboost training and detection algorithm is proposed.The experiment proves that the detection speed of the new algorithm is better than that of the traditional Adaboost algorithm.In the second phase of the face recognition process,the SVM recognition algorithm is studied,and the effect of different features on the recognition effect is very big.In the research process,the LBP feature of the local binary model is improved based on a more commonly used feature,The face recognition algorithm is realized by a feature of image rotation and illumination stabilization,and good results are obtained.The specific work of this paper includes the following parts.1.A multi-resolution Adaboost algorithm is proposed.The traditional Adaboost algorithm is studied,and the traditional Adaboost algorithm is analyzed.The Adaboost algorithm based on multi-resolution training and detection is proposed.The algorithm is used to extract the samples of different scales for human face training.The multi-resolution detector is used to pass through the fixed-scale window to quickly capture the possible position of the face in the image,and then adjust the window in the possible position.Face classifier to detect,you can greatly reduce the number of traversed window when the detection,improve the detection speed.2.A face recognition algorithm based on improved LBP feature is proposed.This paper discusses the characteristics of LBP,discusses the shortcomings of traditional LBP features in face recognition,and uses an improved LBP feature to solve the effects of rotation and illumination on face recognition.The experimental results verify the effect of the algorithm.In this paper,we mainly study the face detection and recognition algorithm,and extend the traditional detection and recognition algorithm.The Adaboost algorithm and the LBP feature are studied in detail,and an effective new method is proposed. |
