Object tracking is an essential part of human-robot interaction, and 2D LiDAR is one of the most popular sensors for wide sensing coverage and excellent ranging accuracy, but its lack of ability to extract semantic information makes it difficult to distinguish objects with close appearance. To address this problem, an object tracking method based on sequence matching between ultra-wide band UWB and LiDAR is proposed. Firstly, the nearest neighbour data association algorithm is used to associate LiDAR clusters and create object trajectories. Then the target is identified and located according to the improved Gaussian similarity between objects' LiDAR trajectories and UWB ranging sequences. Besides, a re-tracking algorithm is designed for the target occlusion case, utilizing UWB ranging information, the target's features, and lost position to find the occluded target. Finally, experiments are performed in a 10 m times 10 m indoor environment with the TurtleBot2 robot platform. The experiment results show that the proposed method can stably track the target carrying the UWB node with an accuracy of 8 centimeters.