论文部分内容阅读
为了充分利用图像中的上下文信息对空间关系进行推理,提出了一种基于产生式模型和区域连接演算(Region Connection Calculus,RCC)的新模型———GM-RCC模型(Generative Model based on RCC),用于合成孔径雷达(SAR)图像的分类研究.首先,通过建立图像金字塔将一幅SAR图像过分割成多尺度的超像素,然后利用层次RCC模型对这些超像素的空间关系进行描述,其中RCC关系的学习和推理都是在产生式模型的框架下进行的.在模型的推理过程中采用了迭代策略以获得更加精细的分类结果.实验选用了极化特征及其他典型特征,并在SAR图像集上进行了实验,实验结果证明了该算法的有效性.
In order to make full use of the contextual information in the image to reason about the spatial relationship, a new model based on production model and Region Connection Calculus (RCC) is proposed. The GM-RCC model (Generative Model based on RCC) , Which is used to study the classification of Synthetic Aperture Radar (SAR) images.Firstly, a SAR image is over-segmented into multi-scale superpixels by building an image pyramid, and then the hierarchical RCC model is used to describe the spatial relationships of these superpixels, The learning and reasoning of RCC relations are all carried out in the framework of generative models.Iterative strategies are adopted in the model reasoning to get more detailed classification results.Polarization features and other typical features are selected for experiments and are used in SAR Experiments were performed on the image set, and the experimental results show the effectiveness of the algorithm.