基于CFAR的高分PolSAR影像桥梁自动识别方法

Automatic Bridge Recognition Method in High Resolution PolSAR Images Based on CFAR Detector

桥梁的自动解译具有重要的应用价值,而在影像分辨率为分米级、桥梁场景复杂、桥梁目标较小的复杂情况下,准确地进行桥梁目标的自动识别比较困难。在分析高分辨率SAR(synthetic aperture radar)影像的统计特征和桥梁特征的基础上,提出了一种新的桥梁自动识别方法。首先采用基于Weibull分布的CFAR(constant false alarm rate)算法检测出潜在桥梁目标,然后基于Wishart-H-Alpha分类和形态学处理提取出桥梁场景区域,随后引入霍夫变换并利用桥梁的场景特征、几何特征和散射特征识别出桥梁目标。采用国产机载XSAR数据和美国AIRSAR数据进行验证,结果表明,该识别方法在复杂情况下能够取得令人满意的识别结果,具有较好的适应性。

The automatic recognition of bridges has both civil and military significance. However, in complicated cases when the image resolution is at the decimeter scale. the bridge scenes are messy and the targets small, and automatic recognition will become quite complicated. Thus, we proposed a novel algorithm based on the analysis of the statistical distribution and features of bridge targets in high-resolution SAR images. A CFAR detector locates potential bridge targets based on the Weibull distribution. Scene areas of bridges are extracted and false alarms are removed by utilizing the features of bridges with the help of Hough transformation. Domestic airborne polarimetric SAR data and AIRSAR data illustrate the effectiveness of this method. Results indicate that this algorithm recognizes bridges in complicated cases with high adaptability.