A novel particle filter-based single extended object tracking system is proposed, which addresses the challenges arising from an unknown and changing orientation of the tracked object. To this end, we factorize the overall distribution of the extended object state, and employ Monte Carlo techniques for orientation estimation. A computation scheme for a single value from all particles scheme regarding the periodic nature of the orientation and the geometric representation of the overall state is proposed, and challenges regarding the computation of the measurement likelihood are discussed. Due to the bounded nature of the sampled (one-dimensional) state, convincing results can already be achieved with merely ten particles, ensuring the computational efficiency of the approach. Extensive evaluation is carried out, demonstrating a significant improvement in comparison with state-of-the-art methods for a variety of noise levels and measurement rates.