中文摘要：

选择江苏省沿海辐射沙洲区南部为试验区，以Landsat TM影像为研究对象，详细分析了面状分布的地面控制点（area-distributed ground control points，AGCP）和沿海堤走向的线状分布的地面控制点（;inear-distributed GCP，LGCP）对海岸带遥感影像几何校正精度的影响，对潮间带99个独立的GPS验证点进行了几何校正的精度评价，得到影像无GCP点控制区域几何校正误差的传播规律及残差的空间分布形态，并给出残差随GCP点距离变化的定量参考范围．研究结果表明，在有GCP点控制的区域，应用AGCP法和LGCP法均可获得好的校正精度，而在潮间带无GCP点控制的区域，两种方法的误差均有不同程度的增大，但LGCP法的误差明显小于AGCP法的．在误差传播的方向性上，随着距离的增加，AGCP法在东向的残差呈稳定而尖锐的上升（R^2=0．827），在北向呈变振幅的振荡上升（R^2=0．517）；LGCP法在与控制线垂直的方向上，残差呈缓慢的线性增加（R^2=0．877），在平行方向，距离的远近对残差变化几乎没有影响．在误差传播的数值上，LGCP法在离海堤30km内几何校正的误差小于20m，而AGCP法在相同的距离范围内仅能得到100m的精度．该结论可为海岸带无GCP点区域TM影像几何校正误差的传播规律提供定量参考依据．

英文摘要：

The purpose is to explore the effect of the spatial distribution of ground control points （GCPs） on the accuracy of imagery rectification in a coastal zone. Choosing southern part of the radiative sandy ridge field along the coast of Jiangsu Province, China as the study area, both area-distributed and linearly distributed GCPs （namely,AGCP and LGCP methods） were used to rectify a coastal Landsat TM image. Rectification accuracy was checked against 99 independent points over the intertidal mudflats of no ground con- trol. Results indicated that the root-mean-square error of residuals over these areas is several times larger than its GCPs-measured counterpart in which the rectification accaracy by the two methods was good. From the directivity of spreading error for the image rectification, it was shown that if the GCPs are spatially dispersed over an area, in eastward residuals fluctuate but increase steadily with distance to the source of control （R2 =0. 827）. In northward they fluctuate around 150 m until 15 km, beyond which they rise steadily at a small range of fluctuation. These residuals are less predictable from distance to the source of control than in eastward （R^2 =0. 517）. And if GCPs are distributed along a control line, residuals rise with distance to it linearly and predictably （R^2 =0. 877） in the direction perpendicular to it. In a direction parallel to it,a distance has little impact on rectification residuals. From the value of spreading error, it could be found that the LGCP method could get an accuracy of about 20 m when the distance to a coastal dyke was less than 30 km while at the same time the AGCP method could only afford the accuracy of about 100 m at the similar distance conditions. This conclusion could give a reasonable reference for the errorspreading rule of Landsat TM image rectification in the area of coast or ocean where there existed little or no GCPs.