论文部分内容阅读
目的 SAR图像中固有的相干斑噪声增加了图像分割的困难。为此,提出一种分布式SAR图像分割算法。方法首先假设图像中同质区域内像素满足同一独立的Gamma分布,依此建立SAR图像模型;为了刻画SAR图像中像素的类属性,建立标号场的MRF(Markov Random Field)模型;在Bayesian理论框架下建立图像分割模型;在多主体系统(MAS)框架下,结合MRF模型和遗传算法(GA)模拟分割模型。MAS结构由分割主体和协调主体组成,其中分割主体利用最大期望值(EM)算法估计MRF模型参数,从而实现全局分割;协调主体利用GA实现全局最优。结果为了验证提出方法的有效性,分别对模拟和RADARSAT-I/II SAR图像进行实验,并与EM和RJMCMC算法比较。本文算法的用户精度、产品精度、总精度及kappa系数均高于EM算法。定性和定量分析结果验证了本文算法的鲁棒性和有效性。结论实验结果表明提出的分布式MAS框架下SAR图像分割方法,能够提高分割精度。该方法适用于中高分辨率单极化的SAR图像,且具有很好的抗噪性。
The speckle noise inherent in the target SAR image increases the difficulty of image segmentation. Therefore, a distributed SAR image segmentation algorithm is proposed. Firstly, suppose the pixels in the homogeneous region of the image satisfy the same independent Gamma distribution and establish the SAR image model. In order to characterize the pixels in the SAR image, a Markov Random Field (MRF) model of the label field is established. In the Bayesian theory framework Under the framework of multi-agent system (MAS), combining MRF model and genetic algorithm (GA) to simulate the segmentation model. The MAS structure consists of the main part and the coordinating part. The main part of the MAS estimates the parameters of the MRF model by using the maximum expectation (EM) algorithm, so as to realize global segmentation. The main part of the MAS uses GA to achieve the global optimum. Results In order to verify the effectiveness of the proposed method, the simulation and RADARSAT-I / II SAR images were tested separately and compared with the EM and RJMCMC algorithms. The user accuracy, product accuracy, overall accuracy and kappa coefficient of this algorithm are higher than the EM algorithm. Qualitative and quantitative analysis results verify the robustness and validity of the proposed algorithm. Conclusion The experimental results show that the proposed SAR image segmentation method based on distributed MAS can improve the segmentation accuracy. The method is suitable for medium and high resolution single-polarized SAR images and has good noise immunity.