Honeybees are a popular model for analyzing collective behavior. The BeesBook system was developed to track the behavior of all individuals in the colony over their entire life spans. The bees are marked with a unique bar code-like marker, the comb is recorded, and the images are evaluated automatically to detect, localize, and decode the markers. This thesis focuses on the development of a tracking method i.e., linking detections through time, and a method to determine correct ids for the resulting tracks. As a result, a random forest classifier was found to perform best to decide if two detections should be linked. The features implemented for it are based on id similarity of the detections and a simple motion model. Evaluation results show that the tracking merges about 90% of the tracks correctly. The id assignment returns correct ids for more than 85% of all computed tracks or track fragments.