基于边缘特征的多源影像空间配准精度研究

Research on the Edge-based Registration Accuracy of Multi-source Remote Sensing Imagery

图像配准技术是数字图像处理的关键技术之一。以甘肃省张掖市为研究区,获取ASTER与MODIS两种数据进行空间配准研究,首先以ASTER数据为基准选择同名点对MODIS数据进行空间配准,然后使用两种方式对其配准精度进行评价:第一种方法充分考虑研究区的地表覆盖特征,利用图像分割、边缘提取及图像匹配等技术,以先验知识为基础提取边缘并进行拟合分析,提出基于边缘进行空间配准精度的分析方法;第二种方法借鉴传统配准方式,选择同名点进行空间配准精度分析。研究表明,经过配准后的几何校正精度优于0.5个像元,能够满足精度要求。

Registration is a key technology of digital image processing. In this research, ASTER and MODIS data covering Zhangye city of Gansu province were collected for registration analysis. MODIS images were geometrically registered on corresponding points in light of ASTER data and then two methods were used to evaluate the accuracy. In accordance with the land cover characteristics of the study area, the first method proposed the analytical approach to edge based spatial registration accuracy. Through the use of density segmentation, edge detection and image matching, the edge curve was extracted on the basis of the threshold determined by prior knowledge. The second method, taking advantage of traditional registration approach dealt with the registration of the corresponding points. The results show that the registration geometric accuracy is better than 0.5 pixels, meeting the accuracy require ments.