一种基于视觉显著性的SAR图像弱目标检测算法

An Algorithm of Dim Targets on SAR Image Detection Based on Visual Saliency Model

针对传统的恒虚警率(CFAR)算法应用于SAR图像弱目标检测存在虚警率高的问题,提出一种基于改进型Itti视觉显著性模型的新算法。该算法首先获取SAR图像的局部方差特征图、亮度频率特征图和全局对比度特征图,然后将特征图经高斯模糊和归一化后,将其非线性融合生成原始SAR图像尺度的显著图,最后从显著图中提取视觉显著性区域作为最终的检测结果。仿真结果表明,通过和改进型CFAR、相干CFAR和二维Otsu检测三种算法的检测性能对比,该算法在检测准确率和时间复杂度上均具有良好的性能。

In view of the high false alarm probability when Constant False Alarm Rate (CFAR) algorithm is applied to detect dim targets on SAR image, an improved Itti algorithm is propsed based on visual saliency model. With this algorithm, feature maps of local variance, brightness frequency and global contrast can be acquired firstly. Then, through Gaussian blur and normalization of such feature maps, saliency maps of the same size with the original SAR image are generated through non-linear fusion. Final detection results are extracted from the visual salience region. Through a comparison between our algorithm and other three algorithms, namely, the improved CFAR, coherent CFAR and two-dimensional OTSU algorithm, the proposed one has a better performance with regard to detection accuracy and time complexity can be find.