The development of deep learning has opened up a new frontier for intelligent transportation systems. Vehicles, as the primary subjects within the system, have long been a focal point of research in intelligent driving systems. However, the detection and tracking of vehicles, particularly in the presence of occlusions caused by other vehicles or structures, often lead to the loss of target vehicle IDs. Therefore, ensuring the consistency of vehicle IDs before and after occlusions has become an urgent problem to be addressed. In response to this challenge, this paper proposes an improved vehicle tracking algorithm based on the advantages of the Byte Track and Bot Sort algorithms. This algorithm combines the fast-processing speed of the Byte Track algorithm with the high precision of the Bot Sort algorithm for vehicle tracking. Additionally, we have restructured the eight dimensions of the state space of the Kalman filter. Through experimental analysis, the proposed tracking algorithm demonstrates a significant reduction in ID conversion occurrences, with an achieved MOTA value of 82.4% and an FPS value of 58. The overall model better addresses occlusion issues during vehicle tracking, meeting the requirements for real-time tracking.