Abstract: To address the problem of poor target tracking accuracy and robustness in the crowded cover scene, we propose a joint feature-space pyramid and adaptive feature update for multi-object tracking. Firstly, a spatial pyramid pooling module combined with channel attention mechanism is introduced into the feature pyramid network, and a feature extraction network for target tracking based on feature-spatial pyramid is constructed. Secondly, we introduce a parallel person re-identification task branch at the network output to perform feature extraction on the detected objects. Finally, a data association algorithm based on weighted adaptive feature update is designed, which not only improves the accuracy of target tracking, but also enhances the reliability of target association. The MOT16 and MOT17 test sets are used for comprehensive comparison and analysis of the method proposed in this study and the existing representative methods. The experimental results demonstrate the effectiveness of the proposed method and its capacity to significantly improve the accuracy and robustness of multi-object tracking.