Proportional Similarity-Based Openmax Classifier for Open Set Recognition in SAR Images

Most of the existing Non-Cooperative Target Recognition (NCTR) systems follow “closed world” assumption, i.e., they only work with what was previously observed. Nevertheless, real world is relatively “open” in sense that knowledge environment incomplete. Therefore, unknown targets can feed recognition system at any time while it operational. Addressing this issue, Openmax classifier has been recently proposed optical domain to make convolutional neural networks (CNN) able reject targets. There are some fundamental limitations end up two potential errors: (1) rejecting a known target and (2) classifying an target. In paper, we propose new increase robustness accuracy. The classifier, which inspired by based on proportional similarity between test image different training classes. We evaluate our method radar images man-made from Moving Stationary Acquisition (MSTAR) dataset. Moreover, more in-depth discussion hyper-parameters detailed description functioning given.