Vision-aware air-ground cooperative target localization for UAV and UGV

In the moving target localization, due to the influence of image distortion on unmanned aerial vehicle (UAV) and the limitation of field of vision on unmanned ground vehicle (UGV), the localization accuracy of single domain is low. Therefore, the air-ground cooperation has become a new trend and inevitable choice for the target localization. In this context, the accurate recognition and organization cooperation are the essential but challenging tasks. This study devotes to improving the capability of air-ground cooperative target localization by the multi-feature fusion recognition in the outdoor environment. To be specific, a novel cooperative architecture is designed to control the observation locations of the UAV and the UGV throughout the localization process. For the heterogeneity of Unmanned Vehicle (UV), we creatively introduce the target recognition rate of the UGV as decision factor and use the target motion property to guarantee the accuracy of the target localization. Based on this architecture, we propose a BidiRectional feedbAck VErification algorithm, BRAVE, whose includes multi-layer decision to ensure that each UV can be in the optimal location for positioning the target, the proposed BRAVE skillfully integrates both the target recognition and cooperation decision. To demonstrate the superiority of our method, we devise several groups of experiments to analyze the performance of BRAVE in the outdoor environment. The experimental results demonstrate that the localization error of BRAVE is only 1.61 m. Moreover, the accuracy is also improved by 0.5 m compared with the air-air cooperation.