ISAR图像的特征提取及应用研究

发布时间：2019-01-30 20:14

【摘要】：逆合成孔径雷达(Inverse Synthetic Aperture Radar,ISAR)技术通常利用微波数据获取高分辨率的雷达图像,其可以对诸如卫星、太空站、飞行器、船舰等空间和空天目标进行远距离、全天时和全天候的二维成像,并对目标进行探测与识别。由于ISAR成像技术不断地普及和深化,其应用范围也从最初的单一成像拓展为对目标部件、运动参数等更为细致的研究。当今,通过获取兴趣目标及其部件的几何结构、散射特性、运动状态与趋势等信息,对后续分析目标的有效载荷、类型识别以及部件分类等提供理论支撑与实践参考价值。本论文利用二维ISAR成像数据和图像为数字资源,以目标的特征提取、参数估计、部件结构融合和兴趣目标部件的检测为主线,对ISAR目标在几何尺寸、形态结构和部件信息方面进行了研究。本文以如下内容展开:第一章介绍了本文的研究背景和目的,介绍了ISAR成像技术的发展过程和利用ISAR图像特征进行相关处理的概述,具体说明了本文主体部分工作安排。第二章介绍了本文工作中所采用的特征点提取方式。通过模型分析并构建目标运动的代价函数,并利用评价准则获取高精度的相邻子孔径ISAR图像之间有效旋转角度,从而获取有效的方位向分辨率。之后,对目标进行定标,分析各部分结构的尺寸大小。同时,利用估计的参数,通过消除相位项方法,从而提高ISAR图像整体的聚焦效果。第三章介绍了通过短相干处理时间下不同子孔径之间目标结构的分析、几何形态的变化和运动参数的估计,完成在长相干处理时间内的不同姿态角度下目标结构融合,以获取目标结构较为完善的ISAR图像,从而增强目标的可识别度与信息可分析性。首先,利用参数估计获取序列子孔径下的运动参数和几何形变尺寸,并以此对相应的ISAR图像进行几何尺寸校正和姿态统一化。之后,利用有效的数据融合方法对不同的结构进行能量阶平衡,从而获取清晰完善目标图像。第四章介绍了利用ISAR图像进行目标局部兴趣部件的检测和识别。其利用目标的整体特征信息与目标部件库下的局部特征信息进行区域限定,通过逐步转移方法,获取每次配准下目标特征的概率精度。之后,通过最终的判断准则获取最佳的识别效果。该研究方法主要针对空间目标,其依赖于目标在ISAR图像中清晰的几何结构和相对较高的分辨率。
[Abstract]:Inverse synthetic Aperture Radar (Inverse Synthetic Aperture Radar,ISAR) technology typically uses microwave data to capture high-resolution radar images that can distance space and space targets such as satellites, space stations, aircraft, ships, etc. All-day and all-weather two-dimensional imaging and target detection and recognition. Because of the popularization and deepening of ISAR imaging technology, its application scope is also expanded from the original single imaging to more detailed research on target components, motion parameters and so on. Nowadays, by obtaining the geometric structure, scattering characteristics, motion state and trend of the object of interest and its components, it provides theoretical support and practical reference value for the subsequent analysis of the payload, type identification and component classification of the target. In this paper, 2D ISAR imaging data and images are used as digital resources. The main line of this paper is the feature extraction of target, parameter estimation, component structure fusion and object component detection. The geometric dimensions of ISAR target are analyzed. Morphological structure and component information were studied. The main contents of this paper are as follows: the first chapter introduces the research background and purpose of this paper, introduces the development process of ISAR imaging technology and summarizes the related processing by using ISAR image features, and explains the work arrangement of the main part of this paper in detail. The second chapter introduces the feature point extraction method used in this paper. The cost function of the target motion is analyzed and constructed by the model, and the effective rotation angle between the adjacent sub-aperture ISAR images with high accuracy is obtained by using the evaluation criterion, so as to obtain the effective azimuth resolution. Then, the target is calibrated and the size of each part of the structure is analyzed. At the same time, by using the estimated parameters and eliminating the phase term, the overall focusing effect of ISAR image is improved. The third chapter introduces the target structure fusion under different attitude angles in the dry processing time by analyzing the target structure, changing the geometric shape and estimating the motion parameters under the short coherent processing time. In order to obtain the ISAR image with perfect target structure, the recognition degree and information analyzability of the target can be enhanced. First, the motion parameters and geometric deformation dimensions under the sequence sub-aperture are obtained by parameter estimation, and the geometric dimension correction and attitude unification of the corresponding ISAR images are carried out. After that, the effective data fusion method is used to balance the energy levels of different structures to obtain clear and perfect target images. Chapter 4 introduces the detection and recognition of local interest components using ISAR images. By using the global feature information of the target and the local feature information under the target component library, the probability accuracy of the target feature under each registration can be obtained by the method of gradual transfer. After that, the best recognition effect is obtained by the final judgment criterion. This method is mainly aimed at the spatial target, which depends on the clear geometric structure and relatively high resolution of the target in the ISAR image.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52

【相似文献】