抄録
In this paper we propose a method of automatic object tracking for soccer players based on the best match of objects' attributes, such as location, dominant color and motion vector, to accurately detect and track soccer players especially in the event of object occlusions. Each detected object in the soccer video is assigned attributes that represent the information extracted from the pixel values of the segmented object. The correlations of their attributes are then observed to find the similarity between objects in two consecutive frames. By observing the best match on attribute similarity, the proposed method can reliably track the objects and consistently provide the object with unique identification even during the event of occlusion as shown in the results of the experiment
