Radar high resolution range profile open set recognition of UAV target

In the field of radar target recognition, non-cooperative targets are often encountered in practical applications. In this case, it is necessary to use an open set recognition method to deal with the problem. This paper proposes an open set recognition method based on high resolution range profile (HRRP) of UAV that trains the network through a distance loss and forces the sample feature output by the model to form a close category cluster. Then it fits the boundary of each category according to the distance between the output vectors and corrects the SoftMax score of samples to achieve rejection. This method does not rely on the threshold to effectively distinguish the boundary between the known class and the unknown class. It can identify the known class samples with a high recognition rate under different sample set openness. Additionally, the rejection accuracy of unknown class samples exceeds 95%, which indicates good open set recognition performance.