利用3S技术对梅里雪山地区植被制图的精度检验分析

Accuracy analysis of vegetation mapping for Meili Snow Mountain area,northwest Yunnan,China

在野外考察的基础上 ,应用 3S技术 ,完成了云南西北部梅里雪山地区的 1∶5 0 0 0 0的植被图 .对已完成的植被图通过野外收集的GPS点进行校正 ,GPS样点数的多少依据统计学抽样调查的样本大小计算而得 ,用这些校正样点数建立混淆矩阵进行植被图精度计算 ,最后利用计算成数方差进行检验 .混淆矩阵计算得出植被图总的判对精度即整体精度OA为 84 7% ,利用计算成数方差检验 ,结果表明大部分类型为 90 %以上 .基于 3S技术完成的植被图精度取决于区域面积大小和植被分类等级 ,而利用遥感技术来划分的植被等级与传统的植被分类等级不完全一致 .

The Meili Snow Mountain (28°20′～28°33′N,98°30′～8°52′E) is a very famous mountain in Northwest Yunnan of China by its rich and well protected biodiversity and Tibetan cultural diversity.By applying 3S (RS-Remote Sensing,GIS-Geography Information System,GPS-Global Position System) technology,the 1∶50 000 vegetation map of Meili Snow Mountain area (total about 332 km 2) was drawn out.The vegetation in this area was classified into 18 vegetation types except for stone,glacier and river system.The vegetation map was rectified by applying the GPS points got from the fields.The numbers of GPS points were calculated by the formula of numbers of samples in statistics.313 GPS points were used to rectify the vegetation map.The numbers were fit for the formula of numbers of samples in Statistics.The accuracy and verify of vegetation types distribution in the map was analyzed by building a Probability Error Matrix (PEM) and through the variance analysis.The results indicated that the overall accuracy (OA) of the vegetation map was 84.7%.The accuracy of vegetation map finished by 3S technology was lied on the area of the region and the grade of vegetation class first,but the grade of vegetation class classified by remote sensing technology disaccord with the traditional vegetation class system.The other factors deciding the accuracy of vegetation were the distinguish ability of Remote Sensing image,the accuracy of distinguish,and the numbers of the samples,including vegetation class experts knowledge.