Simulated star maps serve as convenient inputs for the test of a star sensor, whose standardability mostly depends on the centroid precision of the simulated star image, so it is necessary to accomplish systematic err...Simulated star maps serve as convenient inputs for the test of a star sensor, whose standardability mostly depends on the centroid precision of the simulated star image, so it is necessary to accomplish systematic error compensation for the simple Gaussian PSF（or SPSF, in which PSF denotes point spread function）. Firstly, the error mechanism of the SPSF is described, the reason of centroid deviations of the simulated star images based on SPSF lies in the unreasonable sampling positions（the centers of the covered pixels） of the Gaussian probability density function. Then in reference to the IPSF simulated star image spots regarded as ideal ones, and by means of normalization and numerical fitting, the pixel center offset function expressions are got, so the systematic centroid error compensation can be executed simply by substituting the pixel central position with the offset position in the SPSF. Finally, the centroid precision tests are conducted for the three big error cases of Gaussian radius r = 0.5, 0.6, 0.671 pixel, and the centroid accuracy with the compensated SPSF（when r = 0.5） is improved to 2.83 times that of the primitive SPSF, reaching a 0.008 pixel error, an equivalent level of the IPSF. Besides its simplicity, the compensated SPSF further increases both the shape similarity and the centroid precision of simulated star images, which helps to improve the image quality and the standardability of the outputs of an electronic star map simulator（ESS）.展开更多
Based on the target scatterer density, the range-spread target detection of high-resolution radar is addressed in additive non-Gaussian clutter, which is modeled as a spherically invariant random vector. Firstly, for ...Based on the target scatterer density, the range-spread target detection of high-resolution radar is addressed in additive non-Gaussian clutter, which is modeled as a spherically invariant random vector. Firstly, for sparse scatterer density, the detection of target scatterer in each range cell is derived, and then an M/K detector is proposed to detect the whole range-spread target. Se- condly, an integrating detector is devised to detect a range-spread target with dense scatterer density. Finally, to make the best of the advantages of M/K detector and integrating detector, a robust detector based on scatterer density （DBSD） is designed, which can reduce the probable collapsing loss or quantization error ef- fectively. Moreover, the density decision factor of DBSD is also determined. The formula of the false alarm probability is derived for DBSD. It is proved that the DBSD ensures a constant false alarm rate property. Furthermore, the computational results indi- cate that the DBSD is robust to different clutter one-lag correlations and target scatterer densities. It is also shown that the DBSD out- performs the existing scatterer-density-dependent detector.展开更多
文摘Simulated star maps serve as convenient inputs for the test of a star sensor, whose standardability mostly depends on the centroid precision of the simulated star image, so it is necessary to accomplish systematic error compensation for the simple Gaussian PSF（or SPSF, in which PSF denotes point spread function）. Firstly, the error mechanism of the SPSF is described, the reason of centroid deviations of the simulated star images based on SPSF lies in the unreasonable sampling positions（the centers of the covered pixels） of the Gaussian probability density function. Then in reference to the IPSF simulated star image spots regarded as ideal ones, and by means of normalization and numerical fitting, the pixel center offset function expressions are got, so the systematic centroid error compensation can be executed simply by substituting the pixel central position with the offset position in the SPSF. Finally, the centroid precision tests are conducted for the three big error cases of Gaussian radius r = 0.5, 0.6, 0.671 pixel, and the centroid accuracy with the compensated SPSF（when r = 0.5） is improved to 2.83 times that of the primitive SPSF, reaching a 0.008 pixel error, an equivalent level of the IPSF. Besides its simplicity, the compensated SPSF further increases both the shape similarity and the centroid precision of simulated star images, which helps to improve the image quality and the standardability of the outputs of an electronic star map simulator（ESS）.
基金supported by the National Natural Science Foundation of China (61102166)the Scientific Research Foundation of Naval Aeronautical and Astronautical University for Young Scholars (HY2012)
文摘Based on the target scatterer density, the range-spread target detection of high-resolution radar is addressed in additive non-Gaussian clutter, which is modeled as a spherically invariant random vector. Firstly, for sparse scatterer density, the detection of target scatterer in each range cell is derived, and then an M/K detector is proposed to detect the whole range-spread target. Se- condly, an integrating detector is devised to detect a range-spread target with dense scatterer density. Finally, to make the best of the advantages of M/K detector and integrating detector, a robust detector based on scatterer density （DBSD） is designed, which can reduce the probable collapsing loss or quantization error ef- fectively. Moreover, the density decision factor of DBSD is also determined. The formula of the false alarm probability is derived for DBSD. It is proved that the DBSD ensures a constant false alarm rate property. Furthermore, the computational results indi- cate that the DBSD is robust to different clutter one-lag correlations and target scatterer densities. It is also shown that the DBSD out- performs the existing scatterer-density-dependent detector.
文摘以动态数据驱动技术为基础,通过对林火蔓延模拟精度验证方法和模拟误差的分析,确定模拟误差修正参数及其计算方法,并通过神经网络技术自动生成模拟误差修正参数,从而实现林火蔓延模型模拟误差的在线自适应修正。以王正非林火蔓延模型为例,采用历史记录火场数据对模拟误差的自适应修正过程进行验证试验,结果表明,在预测的16条记录中,有14条与计算结果误差小于预定的0.20 m.min-1,2条误差超过0.20 m.
