Research On Face Real-time Tracking Algorithm Based On Particle Filter And Multi-feature Fusion

With the development of cloud computing and the Internet of Things,face recognition has become an important field in modern research.Face recognition has the advantages of difficulty in copying,complexity of stealing,and simplicity.Face recognition has become an important means of tracking and identification in information security based on today's big data.Face recognition is mainly divided into three aspects: the acquisition of face information,the acquisition of basic face features,and the matching of feature information after obtaining the face.However,in the process of obtaining facial features,it is susceptible to factors such as changes in lighting,the speed of video operation,and object occlusion,resulting in inaccurate final tracking accuracy.Therefore,how to quickly and accurately detect that a human face has become a human face Important research process of detection.It is based on the above research background that we have conducted further research on face recognition and proposed a fast and accurate face detection method.This thesis proposes a multi-feature fusion algorithm that adaptively adjusts the target tracking window and adaptively updates the template's particle filter tracking framework.In the first step,we extract the edge features,color features,and Texture features.In addition,the weighted method is used to weight the histograms of color,edge,and texture features.In the second part,in order to improve the rate of face detection,we use the integrated histogram method to simplify the complex steps of particle calculation and greatly Increasing the operation speed saves operation time.In the third step,we adjust the size of the tracking window in real time according to the change in the average distance from the center of the particle to the edge to achieve real-time accurate tracking of the face tracking object pair.In the fourth step,the algorithm itself can update the tracking module in real time according to the change of the tracking object,thereby improving the accuracy and precision of the real-time tracking.This paper improves and perfects the above method and tests our method through the face video tracking data set.After comparing the experimental data,we get the following results.The video face is subject to lighting changes,the rate of video operation,And in the complex background such as object occlusion,this paper greatly improves the accuracy of video tracking,simplifies the complexity of particle operations,saves operation time,and has high robustness.