Since the vehicle targets in unmanned aerial vehicle (UAV) images are generally small, the existing neural network calculations will cause the detailed information to be seriously lost, which leads to the poor effects of existing recognition algorithms. Here we improved one of the most representative algorithms in the field of target recognition refer to the characteristics of UAV image datasets. There are two main tasks: first, to improve the Faster R-CNN network structure so that it can supplement the detailed information of small targets; second, to verify that it has higher network accuracy, model discrimination and robustness through experiments.