高分辨率光学卫星影像高精度定位技术与实践

High-precision Orientation Algorithms of High-resolution Satellite Imagery

卫星影像高精度定位技术的核心是建立一套适合其成像特点的数学模型和解算方法。在分析严密几何成像模型和通用数学成像模型的基础上,提出了无需了解具体卫星平台、传感器结构和检校参数,具有明确几何意义模型参数,可通过RPC参数计算得到的,理论上适用于各种光学卫星影像几何处理的抽象几何成像模型;并介绍了基于该模型的自检校区域网平差及其在卫星影像高精度定位中的实际应用。为了满足区域网平差对连接点数量和分布的要求,探讨了基于SIFT特征点与角特征点相结合的连接点自动提取算法以及获取高精度同名点像点坐标的方法。针对实际应用中卫星影像通常存在明显系统误差的问题,介绍了自检校区域网平差过程中的3种系统误差在轨检校和补偿方法,并通过实际数据验证了综合使用这3种方法对系统误差检校和补偿改正的有效性。为提高卫星影像无控定位精度,研究了在卫星影像区域网平差中使用SRTM作为控制信息以提高平面和高程无控定位精度的技术和方法。实验数据表明,使用SRTM作为控制信息,单景资源三号立体影像的无控定位精度可达平面5.6 m,高程2.4 m。以一个约18.4万km^2的区域为实验区,介绍了多时相多次覆盖大区域资源三号立体卫星影像无控自由网的整体平差,外业精度检查表明,其平面精度为5.42 m,高程精度为2.85 m。

The geometric sensor model and precise orientation of high-resolution satellite imagery is a fundamental step of its photogrammetric processing. They become crucial in multi-source or large-block data fusion and analysis. We provided a critical review of approaches used for geometric modeling of optical pushbroom sensors in this paper. And then, we proposed a generic sensor model, which was applicable to the geometric processing of various linear-array CCD sensors without using their specific structure knowledge or calibration parameters. We presented its application for on-orbit self-calibration of sensors and block adjustment by using DEM to improve positioning accuracy without any ground control points of multi-source satellite imagery, and proved the viability and good accuracy of the approach by extensive experiments and feasibility tests.