基于深度学习的机载对地目标检测方法

In the actual moving scene, due to the small area of the ground target image taken by the UAV camera and the small information content of the ground target, the detection accuracy of the traditional target detection method for small targets is low. In order to effectively improve the accuracy of ground vehicle recognition and tracking, a Faster r-cnn algorithm based on airborne ground target detection and tracking method is proposed in this paper. The region proposal generated by RPN network is optimized, and the embedded system of UAV is utilized to effectively combine software and hardware, so as to realize airborne detection and tracking of ground targets. Experimental results show that this method can significantly improve the accuracy of ground target detection.