基于MRF的极化SAR图像分割方法研究

发布时间：2018-06-17 02:44

本文选题：合成孔径雷达 + 极化　；参考：《合肥工业大学》2014年硕士论文

【摘要】：合成孔径雷达(Synthetic Aperture Radar, SAR)是一种重要的微波成像工具,具有全天时、全天候、高分辨率等诸多优点,获得了广泛的应用。极化SAR能够获得比传统单极化SAR更加丰富的目标散射信息,为图像的正确解译提供了更好的数据支撑。极化SAR图像分割是图像解译中的关键环节,开展极化SAR图像分割研究对提高极化SAR系统的应用水平和极化SAR数据的利用率具有重要的理论意义和实用价值。马尔可夫随机场(Markov Random Field, MRF)是一种常用的描述图像内空间相关性的模型。它将SAR图像分割转化为最大后验概率估计问题,通过对目标函数的不断优化完成图像分割任务。本文根据极化SAR的数据特征,在MRF的框架之内,利用不同的极化SAR数据形式,提出了两种基于区域MRF的极化SAR图像分割方法,研究工作主要包括如下来两个方面：1.研究提出了一种三维区域MRF的SAR图像分割方法。该方法在充分利用单个极化SAR图像相邻区域之间相似性的基础上,进一步融入多个极化SAR图像相同区域之间的一致性,能够抑制单个极化SAR图像易受相干斑噪声和地物信息不全面性的影响,实现SAR图像的准确分割。2.研究提出了一种基于边缘保持的区域WMRF极化SAR图像分割方法。该方法在极化总功率图上引入降低噪声的SRAD滤波算法,同时提取ICOV梯度信息,在此基础上利用分水岭分割合理划分初始区域；将区域MRF与协方差矩阵的Wishart分布相结合,以区域间的差异度动态调节惩罚系数,有效保持图像细节,为实现极化SAR图像的准确分割提供了一个新途径。针对RADARSAT-2和SIR-C获取的极化SAR图像进行了充分的实验,并进行了较为详细的分析。相较于其他基于MRF的SAR图像分割算法,本文所提算法具有明显的分割准确率,在SAR图像解译方面具有可行性与有效性。
[Abstract]:Synthetic Aperture radar (sar) is an important microwave imaging tool, which has many advantages, such as all-day, all-weather, high-resolution and so on, and has been widely used. Polarimetric SAR can obtain more information of target scattering than traditional single polarimetric SAR, which provides better data support for image interpretation. Polarimetric SAR image segmentation is a key link in image interpretation. The research of polarimetric SAR image segmentation is of great theoretical significance and practical value in improving the application level of polarimetric SAR system and the utilization of polarimetric SAR data. Markov Random Field (MRF) is a commonly used model to describe spatial correlation in images. It transforms SAR image segmentation into the problem of maximum posterior probability estimation and accomplishes the task of image segmentation through the continuous optimization of the objective function. According to the data characteristics of polarimetric SAR, two polarimetric SAR image segmentation methods based on regional MRF are proposed by using different polarimetric SAR data forms within the framework of MRF. In this paper, a 3D region MRF segmentation method for SAR images is proposed. On the basis of taking full advantage of the similarity between adjacent regions of a single polarimetric SAR image, the method further integrates the consistency between the same regions of multiple polarimetric SAR images. It can suppress the single polarimetric SAR image easily affected by speckle noise and incomplete ground information, and realize the accurate segmentation of SAR image. In this paper, a region WMRF polarimetric SAR image segmentation method based on edge preserving is proposed. In this method, the noise reduction SRAD filter algorithm is introduced into the total polarimetric power map, and the ICOV gradient information is extracted. On this basis, the initial region is divided reasonably by watershed segmentation, and the region MRF is combined with the Wishart distribution of the covariance matrix. The penalty coefficient is dynamically adjusted by the difference degree between regions, and the details of the image are kept effectively, which provides a new way for the accurate segmentation of polarimetric SAR images. The polarimetric SAR images obtained by RADARSAT-2 and SIR-C are fully tested and analyzed in detail. Compared with other SAR image segmentation algorithms based on MRF, the proposed algorithm has obvious segmentation accuracy and is feasible and effective in SAR image interpretation.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52

【相似文献】