高分辨率SAR图像车辆目标几何特征提取方法

Geometric features extraction from vehicle targets in high-resolution SAR images

几何特征是目标较直观的物理特征,在SAR图像目标鉴别和分类过程中具有重要的应用。针对高分辨率SAR图像车辆目标切片,构建其几何特征提取的流程。首先在区域分割环节实现车辆目标的二值化分割,而后利用二值化图像提取目标几何特征,其中分别基于最小外切矩形方法和Radon变换方法对车辆目标的长宽尺寸进行估计。采用MSTAR的实测车辆目标数据检验了上述几何特征提取算法和几何特征的鉴别性能。

As the most direct physical attribute of targets, geometric features are usually applied to the target discrimination and classification of SAR images. The processing flow for extracting geometric features from vehicle chips in high resolution SAR images is presented. First, the vehicle chip is segmented in region segmentation step to form a binary image on which the geometric features are calculated. And then, the two methods,, minimum enclosing rectangle method and Radon transform method, are introduced to estimate the length and width of vehicle targets. Finally, the real vehicle target SAR image data in MSTAR database are used to test the methods mentioned above and to validate the discriminating performance of these geometric features.