提出一种新的基于恒虚警率(CFARConstantFalseAlarmRate)技术,确定SAR图像中检测船只整体阈值的方法。该方法采用高斯分布(正态分布)作为SAR图像灰度的概率密度函数,由CFAR技术直接导出用于检测船只整体阈值的计算公式,用记数滤波器滤波去除虚警。该算法避免了复杂公式迭代和求解形状参数计算过程,也避免了用二分法寻找阈值的循环解算过程,提高了检测速度。使用XSAR和ERSSAR图像对该算法进行检验,并与其它算法进行比较,结果显示所提出的算法在检测精度和检测速度上都有明显的改进。

A novel method is presented for ship detection in synthetic aperture radar (SAR) images, which is based on the constant false alarm rate (CFAR) technique and considers the probability density function of sea clutter as Gaussian distribution. All possible ship targets are detected using an overall threshold, which is calculated using the analytic formula. Then a statistic filter is used to eliminate the false ship pixels. This method avoids complicated iteration, calculation of shape parameters and dichotomy threshold, and therefore its accuracy and computation speed are improved, which are demonstrated by the results.In the paper, the main ATR techniques for ship detection in SAR images are reviewed, which include the window filter method, self-adapting threshold method, pdf (probability density function) method and PNN (Probability Neural Network) mo-del. A novel method is then presented, which is based on CFAR technique and Gaussian distribution of sea surface clutter. In this method, CFAR operator is given based on Gaussian distribution (normal distribution), and the statistic filter is introduced to eliminate the false ship pixels, finally the framework of the method is described. The X-SAR and ERS SAR images are used for the algorithm test. Parameters such as detection threshold, computation time, number of detected targets and target pixel numbers are chosen as parameters for comparison with other methods. Results and comparison show that the new method proposed in this paper has advantages of high accuracy and computation speed.