Research Of Multi-angle Face Detection And Tracking

Human face has the characteristics of unique identities,accessibility,and non-invasion.It has become an important method to confirm and track suspects by video analysis.As the majority of the existing face detection and tracking algorithms have difficulty in practical application,such as large angle resulting in missed detection,so it has become a key step to improve the adaptability of angle and is becoming a hot spot.The method of multi-angle face detection and tracking is researched in this paper,including key technologies such as face feature extraction,multi-angle face detection and tracking,designing and implementing the software for multi-angle face detection and tracking.The specific works are as follows:In terms of face feature extraction,the method based on Haar-like,MB-LBP,HOG is researched and implemented,and three AdaBoost classifiers have been constructed.In terms of multi-angle face detection,the method based on multi-feature fusion is proposed.The method combines three kinds of classifiers and outputs face regions by region classification,fusion weight calculation,weighted voting,and threshold decision.The results of simulation experiments show that,the proposed method has better performance than the single feature method,and the detection rate is 92.63% and false rate is 3.15%.In terms of multi-angle face tracking,the method of Camshift tracking based on multifeature fusion and the method of KCF tracking based on multi-feature fusion are proposed.These two methods regard the face region detected by multi-feature fusion as the initial tracking search window.The results of simulation experiments show that,these two methods can track video sequences from PC camera in real time,and when the method of KCF tracking tracks video sequences from video database,the center distance error is 8.49,the overlap rate is 58.78%,and the tracking time is 55.41 ms.In terms of the design and implementation of the software for multi-angle face detection and tracking,the software is designed and implemented based on Visual Studio 2012 platform,OpenCV library functions,and Qt Creator 5.2.0.The software includes 3 modules,which are the ‘user logging',the ‘AdaBoost training',the ‘multi-angle face detection and tracking',and can realize the function of user permissions limit,classifiers training,multi-angle face of static image detection and multi-angle face of dynamic video sequence tracking.