Abstract:

Because sea clutter in high frequency ground wave radar (HFGWR) echoes is chaotic, a new neural network ensembling method for sea clutter predicting is presented. In phase reconstructed space, the predicting performance and error correlation of all artificial neural networks in the neighborhood of the current sea clutter sample are evaluated so as to adaptively choose and dynamically integrate the gualified part among them for sea clutter predicting. The testing result on the objects detecting echo data of the OSMAR2000 HFGWR system shows the predicting precision of the proposed method is higher than the nonensembling prediction method, so it can be used to improve the objects detecting performance of HFGWR.