Multi-vehicle target tracking method based on deep learning

The invention discloses a multi-vehicle target tracking method based on deep learning, which considers the condition that vehicles are shielded in target tracking, proposes a tracking framework combining a single-target tracker and a multi-target tracker, realizes multi-vehicle target tracking by using deep learning, ensures the real-time performance and accuracy of target tracking, and improves the target tracking efficiency. The method solves the problems of identity jump, resetting and target loss caused by the fact that similar vehicles are difficult to distinguish due to partial shielding when multiple vehicles in front are continuously tracked, realizes multi-vehicle target tracking under two conditions that the vehicles are shielded and are not shielded, reduces errors in the vehicle tracking process, guarantees the vehicle tracking reliability, and improves the vehicle tracking efficiency. And the accuracy of vehicle tracking is improved. 本发明公开了一种基于深度学习的多车辆目标跟踪方法，考虑了目标跟踪中车辆被遮挡的情况，提出单目标追踪器与多目标跟踪器结合的跟踪框