【摘要】 运动车辆的检测和跟踪是智能交通系统的关键技术之一，传统的车辆检测和跟踪方法存在着实时性差，易受背景环境干扰、车辆形态相似导致车辆误检等情况。为了解决这个问题，提出基于YOLOv8和ByteTrack的车辆检测和跟踪算法。在车辆检测阶段，针对模型结构复杂、计算量大等问题，将YOLOv8的骨干网络替换为轻量级的网络MobileNetV3,以减少模型的参数量和计算量，保证车辆检测和跟踪的实时性；针对车辆在摄像头拍摄过程中存在误检的问题，将YOLOv8检测头替换为DyHead动态目标检测头，可以更精确地识别目标车辆。最后采用改进YOLOv8检测算法和ByteTrack跟踪算法结合来完成多目标车辆跟踪，经实验证明，该方法在保证精度几乎不变的情况下参数量降低了45.0%,计算量降低了46.3%,证明了算法的有效性，改进后的模型有较好的实时性与跟踪准确率，满足实际的使用需求。更多还原

【Abstract】 The detection and tracking of moving vehicles is one of the key technologies of intelligent transportation system, and the traditional vehicle detection and tracking methods have poor real-time performance, easy to be disturbed by the background environment, and the similarity of vehicle morphology leads to vehicle misdetection.In order to solve this problem, a vehicle detection and tracking algorithm based on YOLOv8 and ByteTrack is proposed.In the vehicle detection stage, for the problems of complex model structure and large computation, the backbone network of YOLOv8 is replaced by a lightweight network MobileNetV3 to reduce the number of parameters and computation of the model and ensure the real-time vehicle detection and tracking; for the problems of vehicle misdetection during camera shooting, the YOLOv8 detection head is replaced by the DyHead dynamic target detection head, which can recognize the target vehicle more accurately.Finally, the improved YOLOv8 detection algorithm and ByteTrack tracking algorithm are combined to complete the multi-target vehicle tracking, and it is proved by the experiments that the number of parameters is reduced by 45.0% and the amount of computation is reduced by 46.3% while the accuracy is almost unchanged, which proves the effectiveness of the algorithm, and the improved model has a better real-time performance and tracking accuracy, which can satisfy the practical use requirements.更多还原