基于ISAR图像序列的目标三维重构

发布时间：2018-03-22 14:10

本文选题：逆合成孔径雷达　切入点：高斯-伽马整形双窗　出处：《哈尔滨工业大学》2014年硕士论文　论文类型：学位论文

【摘要】：逆合成孔径雷达(Inverse synthetic aperture radar,ISAR)是一种针对移动目标在横向-多普勒域成像的信号处理技术。ISAR在结构特点上,主要体现为目标处于持续运动时,雷达位置却是不变的,这样的目标有船舶、卫星、空间站、飞机以及坦克等。由于ISAR成像不受气候、时间以及距离的限制,因此其应用前景备受研究者们青睐。作为ISAR的一个重要研究领域,基于ISAR图像序列的三维重构将极大的丰富目标物体的分类及识别手段。本文完成了对实测数据的三维重构。充分考虑了实际的ISAR成像过程中图像具备的特点,有针对性的提出了解决方法。针对单基地ISAR的成像机制,阐述了ISAR二维图像序列产生的过程,并建立了目标上稳定、可跟踪的散射点构成的观测矩阵模型,随后讨论了二维图像序列完成三维重构所需的条件。在ISAR成像的过程中,由于受到斑点噪声的污染,使得较为常用的Canny边缘检测器失效。本文采用了基于高斯-伽马整形双窗的边缘检测器,提高了ISAR图像的去噪性能,然后利用基于梯度的角点提取算法,获取帧图像上的角点信息(即目标特征点在像平面的投影点)。跟踪目标散射点点在像平面的投影点,采用基于金字塔的Lucas-Kanade光流算法,在图像序列上对它们进行跟踪,图像金字塔解决了像平面内散射点投影的大运动所造成的局部定位不精确问题,放宽了对相邻图像帧运动间隙过小的要求,Lucas-Kanade光流法用来跟踪图像序列间的特征点,跟踪完成后得到特征点投影坐标便构成了观测矩阵。最终重构结果是目标的散射点结构。本文采用的因子分解法,从观测矩阵中分解出运动矩阵和三维结构矩阵,解决了分解不唯一和非正定的问题,最后得到合乎要求的三维结构矩阵,恢复出目标的立体散射结构。该方法最后应用到了模拟的Hotel转台模型和实测的空间站数据上,对两个目标分别进行了散射重构。
[Abstract]:Inverse synthetic aperture radar (ISAR) is a signal processing technique for imaging moving targets in transverse-Doppler domain. Such targets include ships, satellites, space stations, aircraft and tanks. Because ISAR imaging is not limited by climate, time and distance, its application prospects are favored by researchers. As an important research area of ISAR, 3D reconstruction based on ISAR image sequence will greatly enrich the classification and recognition methods of target objects. In this paper, the 3D reconstruction of measured data is completed, and the characteristics of the images in the actual ISAR imaging process are fully considered. Aiming at the imaging mechanism of monostatic ISAR, the process of ISAR two-dimensional image sequence generation is expounded, and the observation matrix model of stable and traceable scattering points on target is established. After that, the conditions for 3D reconstruction of 2D image sequences are discussed. In the process of ISAR imaging, the images are contaminated by speckle noise. In this paper, the edge detector based on Gauss Gamma shaping window is used to improve the denoising performance of ISAR image, and then the corner extraction algorithm based on gradient is used. The corner information on the frame image is obtained (that is, the projection point of the target feature point in the image plane.) tracking the projection point of the scattering point of the target in the image plane, using the Lucas-Kanade optical flow algorithm based on the pyramid, tracking them on the image sequence. The image pyramid solves the problem of local imprecision caused by the large motion of scattering point projection in the image plane, and relaxes the requirement that the motion gap of adjacent image frame is too small. Lucas-Kanade optical flow method is used to track the feature points between image sequences. After tracking, the projective coordinates of feature points are obtained to form the observation matrix, and the final reconstruction result is the scattering point structure of the target. In this paper, the motion matrix and the three-dimensional structure matrix are decomposed from the observation matrix by the factor decomposition method. The problem of non-unique and non-positive definite decomposition is solved, and the three-dimensional structure matrix is obtained, and the three-dimensional scattering structure of the target is restored. Finally, the method is applied to the simulated Hotel turntable model and the measured space station data. The scattering reconstruction of the two targets is carried out respectively.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52

【相似文献】