A Face Tracking Algorithm Based On Filtering And Face Recognition Detection

Target tracking technology is a very critical technology in the field of computer vision.It is mainly used in important public places such as airports,stations,banks,subways,etc.for security.The tracking accuracy of the pure target tracking algorithm is very low,and the actual tracking effect is not achieved.With the outbreak of big data in recent years,the amount of data has grown exponentially,and there has been an algorithm that does not rely on pure target tracking.The accuracy of target detection algorithms based on deep learning is getting higher and higher.Now the solution to the mainstream target tracking is to use detection-based tracking,that is,to predict the target based on high-precision target detection,so as to achieve accuracy and speed.Tracking.According to this research idea,this paper proposes a fusion algorithm of target tracking,face detection and face recognition,which achieves the advantages of speed and precision.A comparative study of Kalman filtering and particle filtering is performed to compare the performance of the corresponding tracking algorithm.Based on the analysis of the Kalman filtering algorithm flow,the Kalman filtering algorithm and the deep learning target detection algorithm Yolov3-tiny were tested.The algorithm flow of particle filtering is analyzed and the same alg orithm experiment is carried out.It is easy to see from the experimental results that the tracking accuracy of the Kalman filter is not good,and the tracking curve is always oscillating;the tracking accuracy of the particle filter is very high,and the tracking curve is very stable.The face detection algorithm MTCNN based on deep learning is analyzed in detail,and the specific algorithm flow is given and the simulation experiment is carried out.Analysis and experiments show that the face detection alg orithm proposed in 2016 is very effective and achieves real-time detection.The adaptive loss performance index with constraints is proposed,which effectively improves and optimizes the existing face recognition algorithm.Face recognition is an important part of target tracking.The premise of target tracking is to identify the target identity.This paper analyzed Adaptive Face,which was the best performing face recognition algorithm.The results show that although the Adaptive Face algorithm with adaptive margins is greatly improved compared to the Cos Face algorithm with fixed margins,the adaptive learning process algorithm is complex and reduces the performance of the algorithm.Based on this analysis,this paper constrains adaptive margins and achieves better results.In summary,this paper proposes a complete face tracking algorithm,which is run by three algorithms in series.In this paper,the implementation details of the algorithm are given,and the corresponding experiments are carried out to prove the effectiveness and practicability of the algorithm.