Chaotic Behavior Analysis and Multifractal Feature Learning of Specular Reflection Ionosphere Clutter for Target Detection

Because ionosphere condition is dynamically varying in the Earth's middle latitude, the long-range detection performance of operating high-frequency surface wave radar (HFSWR) systems is severely restricted by the presence of ionosphere clutter. The development and performance evaluation of target detection under the F-layer specular reflection ionosphere clutter with concentrated power (Fs-concentrated clutter) will be described in this article. First, this article pays attention to the variable chaotic behavior of Fs-concentrated clutter by using amplitude and phase echo data. Then, this article investigates the multifractal property of the power spectrum between the Fs-concentrated clutter and the mixed echo component. The various fractal characteristics of different types of echo component and Doppler scale are analyzed. Next, the singularity intensity correlation function width, the accumulation area of multifractal correlation spectrum, and asymmetry parameter are extracted as the learning features. These multifractal feature parameters are fed into the improved AlexNet and Resnet networks for distinguishing the types of echo component. Experimental results using real HFSWR echoes datasets have demonstrated that this proposed method can increase the detection probability by about 20% than the state-of-the-art detection methods and classical target detection methods under various signal-to-clutter ratio (SCR) background.