机载模拟POLDER多角度图象配准(英文)

Procedures for Registering POLDER Airborne Simulated Multiangular Images

多角度成象对于获取多角度遥感数据是非常重要的。在对多角度数据进行分析之前，首先要对多角度图象进行配准。经过配准，可以识别出相同的地面目标。在配准过程中，一个主要的困难是由于观察角度的不同、双向反射变化及大气的影响造成的图象灰度变化。如何解决灰度变化的影响关系到图象的配准精度。我们在该文中介绍一种多角度图象配准过程，包括预处理、基于层次结构的图象灰度相关配准。为了去掉大角度造成的灰度变化，我们用改进的模板匹配进行配准。这种方法的优点是与配准模板灰度均值无关，只与灰度变化程度有关。配准过程中也同时考虑到纹理的变化。整个配准过程是沿着观测角度增大的方向进行的序列图象配准，以提高大角度图象配准精度。最后，图象的纠正采用局部自适应几何纠正以获得较满意的结果。我们用法国ＣＮＥＳ提供的ＰＯＬＤＥＲ飞机模拟多角度图象进行配准，并提取了小麦和水稻等植被的多角度信息。

Multiangular imagery is an important way of obtaining multiangular remotelysensed data. Before multiangular data can be used for analysis, it is important to register these data. Through registration, the same area can be identified from multiangular images. One of the problem in multiangular image registration is brightness variation due to differences in viewing angle, bidirectional reflectance variation and atmospheric scattering.In this paper, a series processing of registration for multiangular images will be described. It includes preprocessing and hierarchical brightnessbased registration algorithm. In order to remove the brightness differences between two images with large angle, we use an advanced template matching method which is not related to brightness mean to register images. Texture invariation will be considered. Along the viewing angle increasement direction, sequence registration can improve accuracy of large angle image registration. Finally, registration images were obtained by using local adaptive geometric correction.We use POLDER airborne simulated multiangular images which were provided by CNES of France and have obtained more than 30 differential angle registered data of rice and wheat, etc.