Study On Automatic Geo-registration For Google Earth Images

In recent years, with the development of remote sensing and sensor technology, the resource of remote sensing image data is becoming more and more rich. in order to accomplish image fusion, target extraction, change detection, map updating and high resolution image reconstruction, etc, completing precision image geo-registration quickly and effectively has become the urgent problem in application of remote sensing. The traditional image registration method, finding the ground control points (GCP) manually, is slow, low effective and labor intensive. And it is difficult to find GCPs in sub-pixel level duo to the human visual limitation. Therefore, automatic remote sensing image registration technology was proposed and developed in recent years. And it has become a key image pretreatment technology.The Google Earth can provide free high spatial resolution remote sensing images and it updates images frequently. In this study, remote sensing images downloaded from Google Earth were used for geo-registration experiment, and1:10000DOM images were taken as the reference images. According to the terrain of study area, four experiment areas were selected, one in plain area, two in hill area and one in mountain area. For hill and mountain areas, orthographic correction was studied with different type of DEM except for the polynomial correction experiment. According to the study, the RMSE of correction for plain and hill areas were less than2.5meters, and the RMSE value was larger in mountain area. Compared with the RMSE value of validation points for polynomial correction, the RMSE value was smaller for orthographic correction with all kinds of DEM. And the RMSE value for geo-registration was SRTM DEM> Aster DEM>1:50000DEM.Therefore, it was found that the Google Earth images geo-corrected with Erdas automatic registration technology, the correction accuracy could meet1:10000DOM production requirements in plain, hill, and in mountain area. For mountain and hill areas, it was necessary to use orthographic correction with DEM. The1:50000DEM should be used for orthographic correction if1:50000DEM data was available, and the Aster-DEM data could be used for correction if the1:50000DEM data was not present, and the accuracy of correction might be a bit lower.