ID3 algorithm based object discrimination for multi object tracking

Young-Nam Kim, Hye-Yeon Yu, Moon-Hyun Kim

Abstract: In this paper, we assume that a single pedestrian is a single blob, and define the representative properties of blobs to selection between merge and split of blobs. And we collect attribute information of the blob and make decision tree by using the ID3 algorithm. Next, we experimented for dividing the decision tree by directing the status of merge and split, as a result, it is possible to generate a pedestrian trajectory of the individual.