Binarization for time-frequency images of LPI radar signals based on K-means

Abstract Low probability of intercept radar signal is widely used because it difficult to be intercepted by non-cooperative receivers in electronic warfare. We need binarize the time-frequency images when analyzing LPI signals based on distribution. However, existing binarization algorithms cannot distinguish noise from frequency at low signal-to-noise ratios. In this paper, we propose use K-means algorithm gray signals. F1-score comprehensively consider effect binarization. Based linear modulation signals, simulation experiments show that proposed exceeds Otsu’s method 40% 80% ratio -6 dB 2 dB. The excellent and has great application prospects.