Abstract: The existing methods for obtaining the pose information of UAV within a cluster based on communication or radar are greatly affected by signal and environmental factors. To address this issue, a target UAV position and attitude estimation method based on active vision is proposed. This method utilized the image, temporal information and kinematic characteristics of the UAV to estimate its pose. Due to the lack of UAV pose-related datasets, a dataset of pose information for UAV (PAI-UAV) was established in this study, which included images of UAVs and corresponding position and attitude information. On the test set of PAI-UAV, the qualified rates of the predicted distance and three attitude angles of the model were 97.6%, 87.9%, 93.3%, and 91.7%, respectively. The effectiveness of the model was verified through onboard experiments.