基于复数据AT特征的SAR图像车辆目标超像素级CFAR方法

A Superpixel-level CFAR Method Based on Complex Data AT Feature for SAR Image Vehicle Target

根据自然地物和人造地物在单通道合成孔径雷达(SAR)复数据上表现出的实部与虚部统计特性差异,文中提出了一种衡量人造目标(AT)可能性的AT特征,并结合该特征设计一种超像素级恒虚警率(AS-CFAR)检测方法,包括三个步骤:首先,利用简单线性迭代聚类超像素分割算法对SAR图像进行预分割;然后,利用SAR复数据的AT特征从超像素中找出潜在目标超像素和背景超像素;最后,仅对潜在目标超像素进行CFAR检测,且背景区域的选取也参考其AT特征,最大程度保证了背景区域的均匀性。对MiniSAR数据进行实验证明了文中提出的AS-CFAR算法在检测时间和检测目标像素数上的优势。

According to the statistical characteristics'difference between real and imaginary parts of natural and artificial features on the complex data of single-channel synthetic aperture radar(SAR),a feature to measure the possibility of artificial targets(AT)is.proposed,and a superpixel constant false alarm rate detection method(AS-CFAR)is designed based on this feature,which in-cludes three steps.Firstly,the SAR image is pre-segmented by simple linear iterative clustering algorithm.Secondly,using the AT features of SAR complex data from the superpixels to identify potential target superpixels and background superpixels.Finally,only the potential target superpixels are detected by CFAR,and the selection of background area also refers to the AT features,which ensures the homogeneity of background area to the greatest extent.An experiment on MiniSAR data showed the advantages of the proposed AS-CF AR algorithm in terms of detecting time and target pixels.