论文部分内容阅读
针对目前精度评价尺度单一的问题,提出基于直方变差图的多尺度精度评价方法,分别在像元尺度和亚像元尺度进行土地覆盖数据集精度评价。在像元尺度利用驻点作为采样工具直接评价数据集精度;亚像元尺度上,则利用非严格定义的驻点和驻点直方变差图对不同面积和空间结构的优势类进行精度评价。并以浙江北部典型区域为实验区,Landsat TM/ETM+为参考数据,对UMD、IGBP DISCover、MOD12Q1-2001、GLC2000、GlobCover2009等5种大尺度土地覆盖数据集进行多尺度精度评价实验。结果表明,多尺度精度评价方法能够全面地评价土地覆盖数据集的精度,提供更加丰富的多尺度精度信息。像元尺度精度评价可在一定程度上消除由于参考数据与数据集间的空间匹配造成的误差,评价结果更加客观;亚像元尺度精度评价能有效反映亚像元尺度优势地物面积及空间结构与精度的关系。
Aiming at the single problem of the current accuracy evaluation scale, a multi-scale accuracy evaluation method based on histogram of variation of the histogram is proposed. The accuracy of the land cover data set is evaluated at the pixel scale and the sub-pixel scale respectively. At the pixel scale, the accuracy of the dataset is evaluated directly using the stagnation point as the sampling tool. On the sub-pixel scale, the accuracy of the dominant classes with different area and space structure is evaluated by using non-strictly defined stagnation points and stagnation point histogram. Landsat TM / ETM + is a typical experimental area in the north of Zhejiang Province. Five large-scale land cover datasets including UMD, IGBP DISCover, MOD12Q1-2001, GLC2000 and GlobCover2009 are evaluated with multi-scale accuracy. The results show that the multi-scale accuracy evaluation method can comprehensively evaluate the accuracy of land cover data sets and provide richer multi-scale precision information. The pixel scale accuracy evaluation can eliminate the error caused by the spatial matching between the reference data and the data set to a certain extent, and the evaluation result is more objective. The evaluation of the sub-pixel scale accuracy can effectively reflect the area and the spatial structure of the dominant feature of the sub-pixel scale Relationship with accuracy.