针对图像分类效果与图像的空间分辨率之间的关系尚未得到定量分析的问题,该文以中巴卫星CCD图像对冬小麦的分类结果为参考数据,对空间分辨率与制图精度之间的关系进行理论分析。通过在不同空间分辨率状态下分别计算随着硬分类阈值的取值变化而得到的由漏分误差和错分误差所形成的"帕累托边界",来分析硬分类器所能达到的分类精度的极限。结果表明,在空间分辨率相对较低的前提下,图像整体分类效果取决于被分类像元的纯度在图像中的数量分布状况。研究结论可为制作遥感专题图时如何选取合适的空间分辨率图像提供参考。
For revealing the relationship between the spatial resolution and the hard classification effect,the paper referencing to the classification result of winter wheat from a CCD image of the ChinaBrazil Earth Resources Satellite(CBERS),analyzed the relationship between the spatial resolution and the mapping precision.The"Pareto Boundaries"formed by the omission and commission error pairs for each threshold of pixel purity were calculated at different resolution levels to analyze the upper limits accuracy for any hard classifiers.Results indicated that the hard classification accuracy would depend upon the purity distribution of the land type of interest.The study result could provide a reference for selecting the images with appropriate spatial resolution when making thematic maps.