Small Target Detection Methods In High Resolution Sea Clutter

This paper studies on the target detection technology in high-resolution sea clutter. Tar-get detection in sea clutter are used in a wide range of military and civilian areas with broad prospects. The amplitude statistics in low-resolution sea clutter can be described by a Rayleigh distribution, but it begins to show non-Gaussian in amplitude as the radar resolution raises.Characteristics of high-resolution sea clutter has been widespread concerned.The first target detection technology bases on the the statistical probability of sea clut-ter, K distribution is considered to be the perfect model in the existing model of the statistical distribution, but it is large operation to estimate its parameters and hard to get its analytical so-lution. K distribution is equivalent to Weibull distribution with the constraint of their first order origin moment and second order origin moment is equal respectively in this paper, this equiv-alence remain stable in the case of the higher order origin moments. The model parameters is calculated to analytic solutions with moment estimator. The sea clutter detector in the situation of Weibull is designed and proved to have constant false alarm rate.The second target detection technology bases on the fractal. To avoid this difficulty that sea clutter modeling mechanism is too complex, the target detection is to be naturalized to the anomaly detection of clutter-only by the binary hypothesis test. Three feature is extracted from the time series of sea clutter, which proved to distinguish the range of target and clutter-only effectively. The feature vector extracted from sea clutter-only time series forms a training set with convex hull algorithms. To get the purpose of constant false alarm rate, the convex hull is trained with the criterion of losing maximum volume in each iteration, the area is the optimal decision region when the training is completed. The cell will be clutter-only if the feature vector extracted by the time series at cell under text lie within the decision region, otherwise the cell contains a target. The experimental results of IPIX radar datas show that the superiority of feature detector.