全极化SAR数据反演桥面高度

根据高分辨率SAR图像上建筑区的影像特征, 提出了基于灰度共生矩阵(gray-level cooccurrence Matrix, GLCM)纹理分析的建筑区提取方法, 该方法由初步定位和边界调整2个步骤组成, 均遵循特征计算、基于Bhattacharyya距离的特征选择和KNN分类流程, 所不同的是2个步骤中分别采用了逐块和逐点计算纹理特征的方式以兼顾纹理分析的效率和准确性。文中对不同SAR传感器获取的图像进行了实验。实验结果表明, 选用具有最大Bhattacharyya距离值的3或4个特征可以获得较好的初步定位结果, 建筑区的检测率超过80%, 虚警率低于10%；随着边界调整的进行, 检测到的建筑区边界逐渐接近于真实边界。实验结果验证了该算法的有效性。

Estimation of the bridge height over water using SAR image data

As the rapidly growing of availability of high-resolution urban SAR images, analysis of urban environments using SAR images has become an important task in the field of SAR image interpretation. Built-up areas are the dominant structures of urban environments. Detecting and analyzing built-up areas has attracted more and more attention of researchers interested in urban SAR image interpretation. In this paper we propose a method of detecting built-up areas from high-resolution SAR images using the GLCM (Gray-Level Cooccurrence Matrix) textural analysis. Our method is composed of two stages: initial localization of built-up areas and boundary adjustment. Both stages follow a flow of feature computation, Bhattacharyya-Distance-based feature selection and KNN (K-Nearest Neighbor) classification. The difference is that a block-by-block feature computation manner is used in the first stage while a pixel-by-pixel one is used in the second stage. Experiments are performed on images obtained by different SAR sensors. The results indicate that the best three or four features, which have the highest Bhattacharyya distance, lead to the high performance of initial localization, with detection rate higher than 80% and false alarm rate lower than 10%. With the boundary adjustment is implemented, the detected built-up-area boundaries gradually get close to the real boundaries. The experimental results of different SAR images show that the proposed method for built-up area detection is promising.