To date, radar cross section (RCS) measurements at outdoor ranges require manual inspection by test engineers to check for contamination from natural causes. This is an expensive, time-consuming operation that acts as a bottleneck for collecting RCS measurements at these ranges. Advancements in deep learning (DL) have allowed automation of similar types of manual labor. Unfortunately, such techniques normally require enormous training datasets which are impractical to collect at outdoor RCS ranges. We present a novel deep learning method for detecting organic contamination in RCS data that is robust to differing target geometries and performs well even with as little as one contamination measurement for training. Our test case also demonstrates that this method is even capable of outperforming manual human inspection, by uncovering difficult to find mistakes in the original ground truthing.
[1]
Rob Fergus,et al.
Stochastic Pooling for Regularization of Deep Convolutional Neural Networks
,
2013,
ICLR.
[2]
Matthew D. Zeiler.
ADADELTA: An Adaptive Learning Rate Method
,
2012,
ArXiv.
[3]
Nitish Srivastava,et al.
Dropout: a simple way to prevent neural networks from overfitting
,
2014,
J. Mach. Learn. Res..
[4]
Sergey Ioffe,et al.
Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
,
2015,
ICML.
[5]
John Salvatier,et al.
Theano: A Python framework for fast computation of mathematical expressions
,
2016,
ArXiv.