Abstract: To address the severe performance degradation of multi-object tracking algorithms in UAV aerial video caused by motion blur and inter-object occlusion, this study proposes a multi-object tracking method based on a memory checking mechanism. A memory checking module is designed to recover lost trajectories by leveraging historical trajectory information to assist current-frame detection. This approach significantly improves tracking stability and reduces issues such as identity skipping, trajectory interruption, and target loss. In addition, a lightweight HRNetv2 is adopted as the backbone network for feature extraction, thereby reducing the number of model parameters and improving inference speed. Experimental results show that the proposed algorithm significantly improves the stability of multi-object tracking in UAV aerial video and effectively mitigates trajectory interruption and identity switching.