环境减灾卫星多光谱CCD影像自动几何精纠正与正射校正系统

我国环境减灾光学卫星(HJ)多光谱CCD影像由于具有高时空分辨率、大幅宽等优势,能够提供同一区域不同生长阶段的植被信息,是提取具有生态学意义土地覆盖及开展植被生理生态参量反演的重要数据源之一。然而,目前该数据产品的几何定位不准确及山区地形畸变误差使其难以满足应用需求。高定位精度是遥感影像信息提取、参数反演与应用分析的前提,遥感影像的几何精纠正与正射校正是遥感数据预处理面临的首要问题。在分析国内外卫星影像自动化处理系统研究现状的基础上,结合HJ CCD影像幅宽大的特点,构建了HJ CCD影像的自动几何精纠正与正射校正处理系统。与目前商业软件相比,自动几何精纠正与正射校正处理系统采用了自动化的控制点搜索与过滤方法,能有效提高控制点选取的效率与精度。同时,结合DEM数据,系统自动拟合卫星飞行路径并纠正由偏离星下观测导致的山体位移。系统应用结果表明,自动几何精纠正和正射校正系统能够有效的提高处理效率,节省人力和时间成本。该系统已被成功应用于全国生态十年(2000—2010年)变化遥感调查与评估专项土地覆盖遥感监测的环境减灾卫星多光谱遥感影像预处理工作。

Auto-registration and orthorectification system for the HJ-1A/B CCD images

Time-series remote sensing data can provide vegetation information in different stages in one growing period, which is very helpful for the understanding of the dynamic of land cover. As the basic data source of ChinaCover, the multi-spectral CCD images of HJ constellation have the same 30m spatial resolution and visible to near infrared band as other earth resource satellite. It can provide 2 days temporal resolution for there are two satellites named HJ-1A and HJ-1B with 180 degree phase differences and each satellite has 4 days temporal resolution. It is an important data source for the ecological land cover classification and eco-parameter inversion. However, incorrect geolocation caused by the satellite motion along the roll, pitch, yaw direction and the pixel displacement caused by the topographical variations at the off-nadir viewing angle are the bottlenecks when using the massive amount of HJ data. High precision space coordinates is the basic information to guarantee the information extraction, geostatistics analysis and to combine images with other geographic information. The geo-registration and orthorectification is the first issue facing in the pre-processing procedure for the remote sensing images. The newly launched small constellation of environmental and disaster mitigation (HJ-1A/B) are developed for the frequent natural disasters and environmental problems in China. It has been widely applied in the monitoring of the natural disasters, evaluating of the healthy condition of eco-environment and retrieval of surface parameters for its high spatio-temporal resolution and wide coverage. In order to satisfy the immense data processing requirements, this study constructed an auto-registration and orthorectification system for the HJ-1A/B CCD images according to the data characters. Compared with the general commercial software, the system of this study adopted an effectual method to kick out bad tie points to improve the efficiency and accuracy for the selection of tie points. At the same time, this system can fit the nadir tracking path for satellite to correct the pixel displacement caused by off-nadir viewing of topography. The accuracy of this system is evaluated from the validation points which automatically searched from the result and base image, and the validation points are grouped into four sub-regions in order to make a more objective validation for each region. According to statistics of the control points and validation points, the mean error before registration is 188.4m, and standard deviation error is 144.55m. After registration and orthorectification, the mean error and standard deviation error decreased to 40.59m and 21.39m separately. The systematic errors in the original images have been successfully eliminated in the final orthorectified results according to the comparison of spatial error distribution of validation points. The orbit operation rules for HJ1A/B will introduce the distortion on the geometric and spectral characters, and will further influence the time-series images construction because of the reduced overlay images in one revisit cycle. Cloud contamination is the main factors affected the tile points search and the coupled cloud detection algorithm and geometric correction algorithm will increase the geometric correction accuracy for cloudy satellite images and make them more useful. This system has been extensively used in the pre-processing procedure of HJ-1A/B CCD images for the LUCC mapping for the remote sensing survey and assessment project of the National Ecological Environment Decade of Change (2000-2010).