3D object tracker for sports events

Abstract

Objectives. Sports events are currently among the most promising areas for the application of tracking systems. In most cases, such systems are designed to track moving objects in a two-dimensional plane, e.g., players on the field, as well as to identify them by various features. However, as new sports such as drone racing are developed, the problem of determining the position of an object in a three-dimensional coordinate system becomes relevant. The aim of the present work was to develop algorithms and software for a method to perform 3D tracking of moving objects, regardless of the data segmentation technique, and to test this method to estimate the tracking quality.Methods. A method for matching information on the speed and position of objects was selected based on a review and analysis of contemporary tracking methods.Results. The structure of a set of algorithms comprising software for a moving-object tracker for sports events is proposed. Experimental studies were performed on the publicly available APIDIS dataset, where a MOTA metric of 0.858 was obtained. The flight of an FPV quadcopter along a track was also tracked according to the proposed dataset; the 3D path of the drone flight was reconstructed using the tracker data.Conclusions. The results of the experimental studies, which demonstrated the feasibility of using the proposed method to track a quadcopter flight trajectory in a three-dimensional world coordinate system, is also showed that the method is suitable for tracking objects at sports events.