使用测地线活动轮廓模型的合成孔径雷达图像分割方法  

Synthetic Aperture Radar Image Segmentation in Geodesic Active Contour Model

作　　者：孟颖慧 潘杨[1] 朱磊[1] 冯于珍 MENG Ying-hui;PAN Yang;ZHU Lei;FENG Yu-zhen(Electronic Information College,Xi an Polytechnic University,Xi an 710048,China)

机构地区：[1]西安工程大学电子信息学院,西安710048

出　　处：《科学技术与工程》2020年第20期8310-8315,共6页Science Technology and Engineering

基　　金：国家自然科学基金(61971339);陕西省重点研发计划(2019GY-113);西安市科技局创新引导计划(201805030YD8CG14(6))。

摘　　要：为解决经典的测地线活动轮廓(geodesic active contour, GAC)模型在合成孔径雷达(synthetic aperture radar,SAR)图像上不适用的问题,在原GAC模型的基础上提出一种新的分割方法。首先在ROEWA算子进行边缘检测的同时使用Frost滤波器对图像进行负指数加权,形成边滤波边检测的模式,提高了算子的边缘检测与定位能力;然后使用经验阈值对ROEWA强度进行二值化,提高了曲线演化的速度与精确度;最终使用二值化的改进后ROEWA算子替代原GAC模型中的全局梯度项作为边缘指示函数引导图像分割。实验结果表明,针对仿真SAR图像及真实河流SAR图像,分割效果与参数指标都有明显提升。To solve the problem in the classic geodesic active contour(GAC) model that was not suitable for synthetic aperture radar(SAR) image, a new segmentation method was proposed based on the original GAC model. First, when the ROEWA(ratio of exponentially weighted averages) operator was used for edge detection, Frost filter was applied for negative exponential weighting of image to form a mode to detect the edge during filtering, which improved the edge detection and positioning ability of operator. Secondly, an experience threshold was used to binarize the ROEWA intensity, which improved the speed and accuracy of curve evolution. Finally, the improved the binarized ROEWA operator was employed as the edge indicator function to replace the global gradient term in the original GAC model and guide the image segmentation. The new method was testified on real river SAR images. The results show that the segmentation effect and the parameters of the simulated SAR images are significantly improved.

关 键 词：合成孔径雷达(SAR) 测地线活动轮廓模型 ROEWA算子 Frost加权 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]