To obtain a larger aperture for improving parameter estimation performance, we present a novel polarization sensitive array with dual-polarization synthetic nested structure, and then propose a low-complexity method for direction of arrival (DOA) and polarization estimation with two steps. There is a systematic procedure to determine the number and positions of antennas in each single-polarization subarray used for the synthetic nested array, in order to achieve more degrees of freedom (DOFs) and larger aperture for difference coarray with limited physical antennas. The increased DOFs of cross- and self-difference coarrays are quantitatively provided. To handle underdetermined parameter estimation with the resulting DOFs, the proposed method first estimates initial DOA and polarization state by using root-MUSIC and oblique projection operator, based on hole-free cross-difference coarray from cross-correlation matrix between subarrays. In the second step, holes in self-difference coarray, which is from autocorrelation matrix of single-polarization subarray, are filled with initial estimation. By exploiting both hole-free self- and cross-difference coarrays, an efficient search-based algorithm with initially estimated parameters is explored for refined estimation. The Cramér–Rao bound for the array and computational complexity of the method are provided. Simulation results are given to demonstrate the superior performance of the proposed array and method.

Over the past few years, convolutional neural network (CNN) has been broadly adopted in remote sensing (RS) imagery processing areas due to its impressive capabilities in feature extraction. Nevertheless, it is still a challenge for CNN-based hyperspectral image (HSI) classification methods to extract more effective spectral-spatial features considering all spectral bands. Driven by this issue, we propose a novel approach to cope with the HSI classification task, referring to the multilevel joint feature extraction network. The proposed network makes full use of the information on each channel of HSI and transforms it into valid channel-wised spatial features through a designed convolution process. Moreover, these feature maps form global attention details to guide the extraction of spectral-spatial features, which are taken to the next level for further feature mining. Then, the features obtained at different levels are integrated for ground object classification. In contrast with several state-of-the-art HSI classification methods on four public datasets, experimental results demonstrate the effectiveness and remarkable feature extraction capability of our proposed approach.