IEEE Transactions on Aerospace and Electronic Systems · 2022 · 38 citations · 29 references
ComputingComputer ScienceArtificial IntelligenceComputer VisionMachine LearningDeep LearningDigital Signal ProcessingPattern RecognitionRadar Signal ProcessingRadar SystemsSignal ProcessingAutocorrelation FunctionElectronic IntelligenceRadar ImagingRadar SensingSignalsSituational Awareness SystemsSupport MeasuresTransform
Passive systems, such as electronic intelligence and electronic support measures systems, aim to extract necessary information from the received radar signals for situational awareness. To achieve this, the system must first deinterleave the radar signals simultaneously coming from different emitters, so that the pulse repetition interval (PRI) patterns will be revealed for further analysis and identification purposes. PRI transform is a well-known deinterleaving method that utilizes the complex autocorrelation function. There are two main versions of the method. The initial version detects only constant PRI schemes, while the second modified version is capable of detecting varying PRI schemes as well. Miss detection of varying PRI patterns is the drawback for the first version, while producing harmonics, especially at high PRI levels, is the disadvantage of the second one. To alleviate these problems, we propose an image segmentation method based on deep learning. The developed preprocessing step uses both versions of the PRI transform outputs to generate 2-D time–PRI images of the collected radar emissions, so that constant and varying PRI patterns are revealed. The images are concatenated and fed to the proposed network, which uses a practicable U-Net structure. The output of the network directly estimates the PRI levels of the existing radars and the time duration of the transmission jointly. In addition to qualitative and quantitative experiments on the synthetic datasets, qualitative experiments are conducted on real measurements, in which we demonstrate that the proposed method effectively utilizes PRI transform in the preprocessing step and outperforms both versions of the PRI transform in terms of accuracy, Jaccard index, structural similarity, and PRI estimation error metrics.
29
Adam: A Method for Stochastic Optimization
Diederik P. Kingma, Jimmy Ba · UvA-DARE (University of Amsterdam) · 2014 · 84.8K citations · Full text
Image quality assessment: from error visibility to structural similarity
Zhou Wang, Alan C. Bovik, Hamid R. Sheikh et al. · IEEE Transactions on Image Processing · 2004 · 56K citations
Fully convolutional networks for semantic segmentation
Jonathan Long, Evan Shelhamer, Trevor Darrell · 2015 · 36.9K citations
Learning Deconvolution Network for Semantic Segmentation
Hyeonwoo Noh, Seunghoon Hong, Bohyung Han · 2015 · 4K citations