An algorithm of object segmentation in video sequences by methods of fuzzy logic
Abstract
The paper concerns an algorithm of automatic segmentation for multiple objects of interest on the basis of fuzzy logic, taking into account the reliability of moving vectors. We study the attributes and membership functions for blocks and clusters at the initial segmentation stage in the case of the unknown number of clusters. Methods of finding the object characteristics, such as gravity centers, form parameters, movement models, are given. The attributes and membership functions for blocks and objects during tracking are determined. Resolution for situations of occlusion (overlapping objects) and object identification after tracking failure are shown.