Ground moving target detection using two high-resolution aerial images: theory and methods

In drones, high-resolution aerial images can provide a wider field of view and higher spatial resolution than aerial videos. This article investigates the problem of moving target detection using two high-resolution aerial images. We first design a motion quantity based on photogrammetry theory. It is determined by the sensor's intrinsic and extrinsic parameters and target trajectories. Furthermore, the Cramér-Rao lower bound (CRLB) of motion quantity is given, which is verified by simulation experiments. To solve the moving target detection issue, we propose a method based on hypothesis testing, involving testing whether the motion quantity conforms to a hypothesis at a specified confidence level to determine its motion properties. Finally, through extensive experiments, the performance of our method was assessed. The results conclusively showed that when the confidence level is 95%, the accuracy is 97.68%, precision is 95.10%, recall is 82.20%, achieving a good balance between precision and recall.