空间约束混合伽马模型的SAR影像分割算法  

A SAR Image Segmentation Algorithm Based on Spatially Constrained Gamma Mixture Model

作　　者：石雪 SHI Xue(School of Surveying and Mapping Geographic Information,Guilin University of Technology,Guilin,Guangxi 541004,China)

机构地区：[1]桂林理工大学测绘地理信息学院,广西桂林541004

出　　处：《遥感信息》2022年第1期70-79,共10页Remote Sensing Information

基　　金：广西自然科学基金项目(2020JJB150020);桂林理工大学科研启动基金项目(GUTQDJJ2020102)。

摘　　要：为了避免斑点噪声的影响并实现高效且精确的合成孔径雷达(synthetic aperture radar,SAR)影像分割,提出一种结合空间约束混合伽马模型和共轭梯度的SAR影像分割方法。根据SAR影像强度统计特性,采用混合伽马模型建模像素强度统计分布。为了降低SAR影像斑点噪声的影响,利用局部像素类属性定义组份权重,构建空间约束混合伽马模型。考虑到伽马分布自身结构,构建条件期望函数,并采用共轭梯度估计模型参数,以实现高效且精确的SAR影像分割。为了验证所提出算法的分割性能,与对比算法进行分割实验,并定量和定性地评价分割结果。实验结果表明,所提出算法可获得高精度分割结果,且具有较高的分割效率。To avoid the influence of noise and accurately segment synthetic aperture radar(SAR)image,this paper proposed a SAR image segmentation algorithm combining Gamma mixture model with spatial constraint and conjugate gradient. According to the statistical characteristics of intensities in SAR image,Gamma mixture model was used to build the statistical model of SAR image. In order to reduce the influence of noise,the component weight was defined by class attributes of local pixels. Then,Gamma mixture model with spatial constraint was built. Considering the structure of Gamma distribution,conditional expectation function was constructed and conjugate gradient was used to estimate model parameters to achieve efficient and accurate SAR image segmentation. In order to verify the performance of the proposed algorithm,numeral experiments were carried out using the proposed and compared algorithms. The results show that the proposed algorithm has high precision and high efficiency.

关 键 词：SAR影像分割 有限混合模型 空间约束混合伽马模型 条件期望函数 共轭方向 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]