Vehicle target detection is the key to environmental perception in autonomous driving and is of vital importance for maintaining the safe development of intelligent transportation. In view of the problems such as low computational efficiency and susceptibility to interference existing in the current vehicle target detection methods, a vehicle target detection model based on an adaptive feature collection and redistribution module and a single-target detection 11th generation network is proposed. This model enhances the detection performance of the 11th generation YOL single-object detection network by combining the adaptive feature collection and redistribution module, improves the backbone module of the 11th generation YOL single-object detection network by integrating the extruding-excitation attention mechanism and linear detachable convolution, and makes lightweight improvements to the neck module of the network by combining the group mixed wash convolution. To achieve more accurate and efficient vehicle target detection. In the simulation experiment, the vehicle target detection model proposed by the research has detection accuracies of 0.95 and 0.96 for the two target labels of "bus" and "car" respectively, and the recall rates of 0.83 and 0.85 for "bicycle" and "pedestrian" respectively. Both have relatively high detection accuracies. It is indicated that this model can accurately identify vehicle targets of different sizes. The experimental results show that the vehicle target detection model proposed in the research has better detection accuracy and detection rate, and has better generalization ability, indicating that this model can better complete the vehicle target detection task.