Image-Like Situational Difference Mapping (ILSDM)-Based Method for Detecting Anomalies in Electromagnetic Environment

Electromagnetic environment anomaly radiation source detection (EMEARD) can provide a basis for electromagnetic (EM) equipment behavior cognition and battlefield threat assessment. Existing methods mainly focus on generating EM space situations but ignore the research on anomaly signal detection. In this article, the connections between information geometry and performance of sensor networks for anomaly detection are explored. Nonlinear estimation methods based on information geometry can be directly used for anomaly detection. However, Riemann distance as a metric function faces the disadvantages of extensive computation and high complexity. To tackle this issue here, we propose a method based on image-like situation difference mapping, namely, ILSDM, which takes the data-driven strategy to obtain the analytic expression of situation difference mapping distance function (DMDF), and anomaly detection is equivalent to the mapping reconstruction of information geometry. First, the EM space situation is acquired through the distributed sensor network. In addition, to solve the perceptual blind area problem, the sensors’ data are estimated by a spatial autocovariance optimal interpolation algorithm (kriging). Second, we apply the Gaussian mixture model (GMM) to fit the statistical features of each sensing node into probability density function (pdf), and adjacent sensors with similar pdf are fused by the neighbor average fusion method (NAF). Then, based on the information geometry, the fused pdfs are mapped to the Riemannian manifold space as a series of coordinates, and we further propose to use an image-like situation (ILS) to represent the EM space situation. Finally, DMDF measures the information distance between different ILSs to achieve anomaly detection. Numerical simulations and real experiments were conducted to prove the superiority and feasibility of our method for anomaly detection in EM space.