论文部分内容阅读
为了从合成孔径雷达(SAR)遥感图像中高效率地提取线特征、满足目标识别与场景分析等应用的需要,采用曲线结构基元取代通用的直线结构基元,通过像素编组以及曲线拟合提取并连接,使得复杂的全局优化简化为基于连接关系和总长度的聚类和筛选,借助低分辨率图像对高分辨率图像的掩码操作先主后次地提取不同宽度的线特征。应用所提出的方法在实际SAR图像上进行了实验,获得了与观察相一致、具有单像素宽度的线特征二值图像。实验结果表明:所提出的方法可以快速准确地提取出场景中真实的线特征,采用曲线基元形式和编组拟合方法有利于降低问题复杂度、提高处理效率。
In order to efficiently extract line features from SAR images and meet the needs of target recognition and scene analysis applications, the paper adopts the structure of primitives to replace the common linear structure primitives and extracts them by pixel grouping and curve fitting The complex global optimization is simplified to the clustering and filtering based on the connection relation and the total length. The masking operations of the high-resolution image by the low-resolution image are taken first and then the line features of different widths are extracted first. The proposed method is applied to real SAR images, and the line feature binary image with single pixel width is obtained in accordance with the observation. The experimental results show that the proposed method can extract the true line features quickly and accurately. Using the curve primitives and the method of grouping fitting can reduce the complexity of the problem and improve the processing efficiency.