An infrared sensor is a commonly used imaging device. Unmanned aerial vehicles, the most promising moving platform, each play a vital role in their own field, respectively. However, the two devices are seldom combined in automatic ground vehicle detection tasks. Therefore, how to make full use of them—especially in ground vehicle detection based on aerial imagery–has aroused wide academic concern. However, due to the aerial imagery’s low-resolution and the vehicle detection’s complexity, how to extract remarkable features and handle pose variations, view changes as well as surrounding radiation remains a challenge. In fact, these typical abstract features extracted by convolutional neural networks are more recognizable than the engineering features, and those complex conditions involved can be learned and memorized before. In this paper, a novel approach towards ground vehicle detection in aerial infrared images based on a convolutional neural network is proposed. The UAV and the infrared sensor used in this application are firstly introduced. Then, a novel aerial moving platform is built and an aerial infrared vehicle dataset is unprecedentedly constructed. We publicly release this dataset (NPU_CS_UAV_IR_DATA), which can be used for the following research in this field. Next, an end-to-end convolutional neural network is built. With large amounts of recognized features being iteratively learned, a real-time ground vehicle model is constructed. It has the unique ability to detect both the stationary vehicles and moving vehicles in real urban environments. We evaluate the proposed algorithm on some low–resolution aerial infrared images. Experiments on the NPU_CS_UAV_IR_DATA dataset demonstrate that the proposed method is effective and efficient to recognize the ground vehicles. Moreover it can accomplish the task in real-time while achieving superior performances in leak and false alarm ratio.