Influence of Membership Function Type on the Accuracy of Video Tracking Algorithm

The article presents the results of intuitionistic fuzzy sets using in practice, as well as the study of the used belonging and non-belonging functions influence on the result of the video object tracking algorithm work. Fuzzy logic is used for evaluating the degree of assigned correspondences truth between keypoints of the object being followed image and the search window. The algorithm's accuracy estimates are based on synthetically generated data provided with a ground true markup that allows evaluating the degree of calculated correspondences truth. In an effort to obtain objective algorithm estimates, several different metrics are used.