基于合成孔径雷达图像的河流检测方法研究

摘要

针对合成孔径雷达(Synthetic Aperture Radar,SAR)图像中水体的梯度比陆地的梯度要小的特点,提出了一种利用图像梯度信息和Fisher准则检测河流的方法;首先利用梯度算子求出图像的梯度;再对得到的梯度图像运用Fisher准则分割出水体区域;最后利用数学形态学滤除分割图像中存在的大量细小噪声,并运用面积、灰度、方差等特征进一步消除虚警目标,检测出河流;实验结果表明,该方法具有较强的抗相干斑噪声特性,能够自动、快速地检测出SAR图像中河流目标.%According to river feature that water body gradient is less than that of land in SAR (Synthetic Aperture Radar) image, a river detection method based on gradient information and Fisher discriminant is presented. Firstly, gradient operator calculates image gradient, and gradient image is gotten. Secondly, Fisher diseriminant is adopted to segment water body from gradient image. Lastly, river is detected, because much small noise is filtered by morphology method and false alarm targets are excluded according to the featurea of area, grayseale, variance. Experiment results show this method is antinoise and can automatically and rapidly detect river in SAR image.