Acquisition of the players’ trajectories

We have developed three different methods to track the players. Probably the easiest method to implement is GPS tracking. There, each player is equipped with a GPS logger, which determines and stores the positions with a frequency of 5 Hz. Due to a missing communication module, a direct transfer of the data for an online-analysis is not possible. The evaluation must therefore take place afterwards. In addition, the GPS-typical inaccuracies for the player positioning must be expected. However, the advantage of this method is the simple generation of the trajectories, since the stored data of a logger can be easily assigned to a certain player. The second variant is to use cheap (smart) cameras to track the players. There are some advantages and disadvantages compared to the GPS tracking method. The accuracy is significantly higher using the video-based tracking. Depending on the scenario it is in the range of few centimeters. However, due to various obstacles like object occlusions, the assignment of the object detections in the camera images to the respective players is rather challenging. A sophisticated tracking algorithm has to be used to solve this problem. Our hardware setting consists of commercially available smartphones, which are equipped with a sufficiently powerful camera and CPU. The latter even allows a decentralized processing of the images, so that only the detections or the determined player positions have to be transmitted for an online-analysis. The last tracking method is a combination of the two previous ones. The so-called GPS-aided video tracking is intended to bring together the advantages of the two individual methods and thus to eliminate their weaknesses. This means we combine the accuracy of camera-based tracking with the easy assignment of players using the GPS tracks. For this purpose, the data of the camera(s) and the logger(s) are fused using appropriate algorithms. Due to the lack of communication capability of the GPS-Logger, this procedure currently allows only an analysis of the data in a post-processing. For the possible case that the GPS loggers are equipped with a communication unit in the future, the data could be directly merged and analyzed. Depending on the used tracking method, the trajectories are processed either online, so immediately after recording, or in a subsequent analysis phase. In both cases, both the movement data and metadata are stored in a database so that they can later be called up again or used for “long-term analyzes”.