Vehicle Re-identification is designed to locate and match specific vehicles from images or videos captured by a set of non-overlapping cameras. This technology has broad application prospects in intelligent transportation systems, urban security, and autonomous driving. In recent years, methods based on convolutional neural networks have made significant progress in the field of vehicle re-identification. However, existing methods still have shortcomings in dealing with background information interference and attribute feature extraction. To address these issues, we propose a vehicle re-identification method called Detection-Guided Feature Aggregation Network (DFA-Net). DFA-Net first performs vehicle detection and image cropping using YOLO to eliminate interference from non-target regions. Then, it employs two attribute classification networks as attribute feature extractors to extract the color and type features of vehicles, respectively. Finally, it introduces a Spatial Attribute Gating Fusion (SAGF) module to dynamically fuse the global features and attribute features of vehicles through learnable spatial weights. Experiments on the Veri-776 and VehicleID datasets show that DFA-Net has increased the mAP by an average of 8.01% on the Veri-776 dataset and by 9.93% on the VehicleID dataset. These results demonstrate the robustness and superiority of DFA-Net in the task of vehicle re-identification.