Recently, target detection using aerial imagery from Unmanned Aerial vehicles (UAVs) has made notable progress and emerged as a hot study direction. However, the existing target detection methods are based on a closed-world assumption and may ignore valuable unknown information. A crucial challenge in aerial imagery is the capacity to predict known targets with labels in an open-world setting while also continuously learning and predicting unknown classes. Therefore, an Open World object detection method in Aerial Imagery based on Faster R-CNN (OWAI) is present in this paper. Firstly, the aerial image dataset undergoes preprocessing by superimposing region clipping. Secondly, the task of target search is performed to detect potential unknown instances, which utilizes the RPN of Faster R-CNN with class agnosticism to establish a forged annotation of unknown instances. Finally, the task of target classification is completed to distinguish between known and unknown targets and prevent unknown information in the backdrop from being overlooked by creating a feature comparison-based target clustering model and an energy-based classification model. The experimental outcomes on the DOTA dataset illustrate that the approach effectively obtains unknown target information in UAV aerial imagery.