Segmentation And The Incident Angle Effect Correction Of SAR Image

Synthetic aperture radar is an important microwave imaging tool in the field of remote sensing. The obvious incidence angle effect in airborne and wide observing spaceborne SAR image is a crucial factor which affects the precision of SAR image interpretation. The polarimetric SAR is a new type of imaging radar with multi-parameter and multi-channel. It acquires more scattering information of object and significantly enhances the ability of polarimetric SAR image interpretation. The segmentation of polarimetric SAR image is a key procedure in the interpretation of polarimetric SAR image and occupies a special place in image engineering. In this dissertation, the technology both of SAR image incidence angle effect correction and multi-polarimetric SAR image segmentation are studied deeply. The main contribution and contents of this work are as follows:1. In this dissertation, the scattering intensity bias caused by the incidence angle in SAR image is illustrated in highlight. An approximate mathematical expression of the incidence angle bias field in SAR image is deduced on the basis of study on the SAR incidence angle dependence of the terrain backscattering coefficient. This dissertation also presented the basic theory of polarimetric SAR, in which the representation formats of polarization SAR data are summarized and the target polarimetric feature decomposition is briefly introduced. On these bases, the statistical distribution model of polarimetric SAR data and the similarity measurement method of scattering characteristic are systematically investigated.2. This dissertation did a research on the existing method of incidence angle effect correction for SAR image. Then a multi-exponential model based bias field correction methodology is proposed for SAR image. The experiment result indicates that the proposed algorithm is effective in correcting the incidence angle bias field of SAR image. Besides, the proposed algorithm has better correction result than the existing incidence angle correction method of SAR image and it does not need the incidence angle information of the pixels.3. For polarimetric SAR image segmentation, the polarimetric SAR image segmentation method based on MRF (Markov Random Field) model is studied in depth. The Maximum Likelihood (ML) segmentation algorithm and the unsupervised H/αclassification algorithm are also studied. Since the pixel based MRF segmentation algorithm is sensitive to noise and computationally expensive, the region based MRF segmentation algorithm is researched and improved. The similarity of target scattering properties is introduced into the region based MRF segmentation algorithm, so its segmentation result is improved. Segmentation experiment is carried out on the L band full polarimetric data of San Francisco Bay acquired by AIRSAR and the result shows the effectiveness of the proposed algorithm.