A Drogue Detection Method for Vision-Based Autonomous Aerial Refueling

Drogue detection is important for vision-based autonomous aerial refueling. It is a difficult task due to disturbances caused by both the tanker wake vortex and atmospheric turbulence. In this paper, the problem of drogue detection is considered as moving object detection. A method based on multi-scale low rank and sparse decomposition is proposed for drogue detection. Firstly, the image sequences are decomposed by stationary wavelet transform respectively. Then the method based on low rank and sparse decomposition is employed on the low frequency sub-band image sequences of coarsest scale. After obtaining the object, the object is used as object confidence map to feedback the low sub-band image sequences of next fine scale for low rank and sparse decomposition. Thereinto, an alignment method is introduced into the low rank and sparse composition due to the vibration in the drogue video image. This method can overcome the influence of non-structure information, caused by atmospheric or aircrafts. The experimental results show that the proposed algorithm is effective in real autonomous aerial refueling data via multi-scale low rank and sparse decomposition for drogue detection under complex background.