中文摘要：

在超宽带合成孔径雷达叶簇隐蔽目标检测中,传统的UWB SAR图像变化检测方法易受图像灰度值起伏和成像条件变化的影响,致使现有的变化检测算法的性能下降.本文根据人类视觉系统的生理结构和认知特点,提出了一种基于视觉注意机制的叶簇隐蔽目标变化检测算法.该方法使用视觉注意模型,将图像的多尺度特征信息融合为单幅视觉显著图像,并利用图像局部邻域信息和目标的空间相关特性对视觉显著图中视觉注意焦点进行分层筛选和变化检测.实验结果表明：本文中基于视觉注意机制的变化检测方法可以有效检测多时相UWB SAR图像中的叶簇隐蔽目标,较之传统的基于统计原理的变化检测方法,其检测速度更快,且对场景复杂的UWB SAR图像亦具有鲁棒性.

英文摘要：

Traditional statistically-based UWB SAR image change detection is usually limited by large pixel value change between multi-temporal UWB SAR images in foliage-concealed targets detection. In order to deal with this problem,a newUWB SAR foliage target change detection algorithm based on bottom-up visual attention is brought up. In the algorithm,multi-scale image features are combined into a single visual saliency map by using the model of visual attention. Then the image local neighborhood information and spatial correlation are used to improve the performance of detection. Finally,the experimental results showthat the algorithm can detect the foliage-concealed targets between multi-temporal UWB SAR images with less time and it is robust to registration error.