Target Tracking Research Based On Time Regular Items And Multi-Feature Fusion

The core idea of target tracking is to perform training and detection through the feature extraction of the target,and obtain accurate information in the first frame of the video,and then update it according to the predicted information.The rapid development of target tracking has brought a solid foundation,but there is a problem of tracking drift when there is interference such as occlusion or complex background,and the tracking fails due to the influence of noise.Therefore,a target tracking based on temporal regular term and multi-feature fusion is proposed.algorithm.First,the algorithm in this paper introduces a time regular term into the correlation filtering model,uses the eigenvalues to control the similarity of time,and flexibly alleviates the model degradation,so as to achieve the effect of suppressing the negative impact of abnormal data and overfitting in the training process.Secondly,an image classification mechanism is introduced,and the weights of the peak sidelobe ratio and the smoothness constraint are used to realize the adaptive fusion of features.Then,the model is approximated as a linear equality constraint problem,and the(Alternating Direction Method of Multipliers,ADMM)is used to iteratively solve the problem to improve the efficiency of the algorithm.Finally,the Adaboost-SVM classifier is used for re-detection to improve the occlusion judgment ability of the model and improve the tracking efficiency on the basis of ensuring the tracking accuracy.In order to verify the performance of the algorithm in this paper,the target tracking algorithm based on time regular term and multi-feature fusion is compared with the f DSST algorithm on the data set OTB100,the data set Temple Color 128 and the data set La SOT,showing good tracking effect,accuracy and success.The comparison results show that the algorithm in this paper is accurate and robust.The paper has 34 figures,7 tables,and 58 references.