Research On Target Face Tracking Technology Based On Video Sequence

Face detection and tracking technology is of great research value in the field of pattern recognition and computer vision.It can be used in many applications in real life,and it has very closely related to human beings.Under the era of big data in the Internet,national security,intelligent video surveillance,border security,e-commerce and many other fields can see the shadows of face detection,face recognition and face tracking,Face detection and face tracking are the key links in that field.In recent years,it has aroused attention from more and more researchers.by Considering the feature fusion stage in face tracking,this paper design a face detection method based on skin color segmentation and a face tracking method based on Staple algorithm with improved adaptive face features.The details are as follows:(1)By studying a large number literatures of face detection and face tracking,we first introduce the development and current status of face detection and face tracking,sum up the face detection based on prior knowledge and face detection based on statistics.and sum up the face tracking method,there are tracking based on model,tracking based on motion information,tracking based on face local feature,and tracking based on neural network etc in recent years.(2)For the problem of face detection,the method of skin color probability feature is used to describe the person's face.Firstly,in order to correct the color shift,the first frame of the video is subjected to illumination compensation preprocessing to extract the skin color probability features;then,by compareing and filterng in three color spaces,and select the YCbCr color space that separates the luminance and chrominance of the face to highlight the clustering characteristics of the face.Use the a priori feature of the face to retain the qualified face area;finally mark the located area in the original image.(3)Based on the research of correlation-based Staple algorithm,this paper proposes a face tracking algorithm with novel confidence mechanism.Firstly,the position filter is used to extract the HOG features of the face region detected by the face,and the probabilistic model of multi-color space is used to extract.The color features of the face area detected by the face are reserved.The advantage of the HOG feature is that the color change processing is stable,and the probabilistic model of multi-color spacehas a stable processing effect on the spatial change of the face.The fusion of the two features can solve most of the problems encountered in the tracking process,but the target face disappears in the video area or When coincident with other non-target faces,the a priori features extracted from the first frame of the video can be used to track the confidence judgment,and the fusion features are updated in time.A face detection and tracking algorithm that can overcome the movement,disappearance and change of expression of the target face position is designed and implemented.Experiments show that the improved algorithm is much better than the previous ones.