Sentinel-1双极化数据舰船目标几何特性提取

发布时间：2018-06-15 01:40

本文选题：合成孔径雷达(SAR) + Sentienl-　；参考：《科技导报》2017年20期

【摘要】：舰船目标几何特性提取是合成孔径雷达(SAR)图像海上目标检测识别的重要基础。在具有几何真值样本的基础上,通过参数寻优和拟合回归,能够提高几何特性提取的精度,这在Terra SAR-X数据上已有研究。本文考虑Sentinel-1大部分情况下均能提供双极化数据这一特点,探索双极化信息能否进一步提升几何特性提取的精度。基于Open SARShip测试库,首先使用二维度滤波进行图像处理,该图像处理过程中的关键参数使用交叉熵方法进行寻优,在大样本基础上,得到最优参数;之后,在目标几何特性的图像处理提取结果上,综合传感器、环境、目标3方面信息,特别是融合双极化信息,使用多元线性回归模型进行拟合,得到比仅用单极化信息更高的几何特性提取精度,证实了双极化信息的可用性。
[Abstract]:Geometric feature extraction of ship targets is an important basis for marine target detection and recognition in synthetic Aperture Radar (SAR) images. On the basis of geometric true value samples, the precision of geometric feature extraction can be improved by parameter optimization and fitting regression, which has been studied on Terra SAR-X data. In this paper, we consider that Sentinel-1 can provide bipolarization data in most cases, and explore whether bipolarization information can further improve the precision of geometric feature extraction. Based on Open SARShip test library, 2-D filtering is first used to process the image. The key parameters in the image processing process are optimized by cross-entropy method, and the optimal parameters are obtained on the basis of large samples. In the image processing of the geometric characteristics of the target, the information of sensor, environment, target 3, especially the information of double polarization, is synthesized and fitted by multivariate linear regression model. The accuracy of geometric characteristic extraction is higher than that of single polarization information, and the availability of double polarization information is verified.
【作者单位】：上海交通大学智能探测与识别上海市重点实验室;
【基金】：国家自然科学基金重点项目(61331015)
【分类号】：TN957.52

【相似文献】