This study concerns the fractal properties of sea clutter in the power spectrum domain. To overcome the deficiencies of Fourier transform analysis, the power spectrum of the sea clutter is obtained by autoregressive (AR) spectrum estimation. The AR model is a linear predictive model, which estimates the power spectrum of sea clutter form its autocorrelation matrix and has a higher frequency resolution than Fourier analysis. This study concentrates on analysing the fractal property of the power spectrum based on AR spectral estimation and its application on weak target detection. First, fractional Brownian motion is taken as an example to prove the fractal property of the power spectrum. Then, real measured X‐band data is used to verify the fractal property of the power spectrum of sea clutter. Finally, a novel detection method based on AR Hurst exponent is proposed and the factors influencing the fractal properties of power spectrum are analysed. The results show that the Hurst exponent of AR spectrum is effective for weak target detection in sea clutter background. Compared with the existing fractal method and the traditional constant false alarm rate (CFAR) method, the proposed method has a better detection performance.

In this letter, a small target detection method in sea clutter background is proposed based on multiview polarization features and localized one-class support vector machine (LOCSVM). In the proposed method, three views of features that represent the divergence between the small target and the clutter are extracted: 1) the study of the target scattering characteristics shows that the dominant scattering components of the small target cell (TC) are sphere scattering and diplane scattering, while the scattering mechanism of the clutter cell is varied and affected by the sea conditions. Therefore, the relative powers of the sphere and diplane scattering components are extracted based on the polarimetric scattering matrix; 2) Because the target is a definite shaped object, its cell does not exist in the fractal characteristics, while the clutter cell is on the contrary. Therefore, the average Tsallis entropy (TE) is extracted from full-polarization channels to reflect the fractal characteristics of the cell; and 3) As the generalization of the Doppler entropy, because the frequency spectrum of the TC is more congregate, the TE can reflect the Doppler characteristics. Then, in the detection stage, a novel one-class classifier, referred to as LOCSVM, is designed as the detector. LOCSVM is a combination of $k$ -means clustering and OCSVM and can enhance the detection performance of OCSVM. The experimental results on the IPIX data show the effectiveness of the proposed method, especially in the case of short observation time.

Sea surface target detection is a key stage in a typical target detection system and directly influences the performance of the whole system. As an effective discriminator, the one-class support vector machine (OCSVM) has been widely used in target detection. In OCSCM, training samples are first mapped to the hypersphere in the kernel space with the Gaussian kernel function, and then, a linear classification hyperplane is constructed in each cluster to separate target samples from other classes of samples. However, when the distribution of the original data is complex, the transformed data in the kernel space may be nonlinearly separable. In this situation, OCSVM cannot classify the data correctly, because only a linear hyperplane is constructed in the kernel space. To solve this problem, a novel one-class classification algorithm, referred to as ensemble one-class support vector machine (En-OCSVM), is proposed in this paper. En-OCSVM is a hybrid model based on $k$-means clustering and OCSVM. In En-OCSVM, training samples are clustered in the kernel space with the $k$-means clustering algorithm, while a linear decision hyperplane is constructed in each cluster. With the combination of multiple linear classification hyperplanes, a complex nonlinear classification boundary can be achieved in the kernel space. Moreover, the joint optimization of the $k$-means clustering model and OCSVM model is realized in the proposed method, which ensures the linear separability of each cluster. The experimental results based on the synthetic dataset, benchmark datasets, IPIX datasets, and SAR real data demonstrate the better performance of our method over other related methods.

This letter considers the distributed fusion for Labeled Multi-Bernoulli (LMB) filters under the multi-sensor multi-target tracking scenario. In practice, a novel fusion method that combines the label-free strategy with the fusion method based on the Arithmetic Average (AA) is proposed. Firstly, the label-free version of the LMB posterior is obtained. Then, the corresponding Probability Hypothesis Density (PHD) is acquired and fused by the merging-based PHD-AA fusion method. Finally, the labels of the fused distributions are reassigned. The simulation shows that the distributed LMB filter using the proposed fusion method can achieve more accurate multi-target state estimates with relatively low computational burden under clutter and missed detection environment compared to other existing methods.

Weak target detection based on fractal analysis is a hot research topic. The existing methods, using single fractal dimension and Hurst exponent, are not applicable under low signal clutter ratio (SCR) conditions. This study focuses on enhancing the performance of weak target detection by utilising the multifractal properties of sea clutter. The existing global multifractal spectrum based method has limited performance because of its non‐stability. Therefore this study proposes a weak target detection method by taking into account the local difference of multifractal spectrum between sea clutter with and without target. The local‐multifractal spectrum is obtained by adding a rectangle window to the multifractal spectrum, and the local difference is calculated by the local mean square summation algorithm. In addition, the window length, a key parameter for obtaining the local‐multifractal spectrum, is adaptively computed by the slope of the singularity intensity. In comparison with the existing multifractal spectrum based method, the experimental results of real S ‐band and X ‐band sea clutter data show that the proposed method is more stable, and could achieve a better detection performance under low SCR conditions.